+AI Agent Governance
---+Infrastructure Hardening & Isolation
---+AI Agent Limitations
---+Tool & API Invocation Controls
---+Orchestration Protocol Safeguards
---+Data Pipeline & Input Integrity
---+Privileged Role & Delegation Boundaries
---+AI Agent Data Access Restrictions
---+Data Extraction
---+AI Agent Identity & Impersonation Defense
---+AI Agent Logic Integrity
---+Sandboxing AI Agents
---+Prompt Injection Defense
---+Agent Kill Switch / User Control
---+Adversarial & Red Team Testing
---+Self-Modification Controls
---+Purging AI Agent Data
---+Delegation and Chaining Control
---+Behavioral Drift Detection
---+AI Agent Action Authentication & Authorization
---+Transparency & Audit
---+Explainability
---+Ethics, Fairness & Bias Detection
---+Agent Output Integrity & Verification

AI Agent Governance

Description

Mechanisms exist to ensure AI agents are designed, developed and deployed to securely operate under human oversight.

Possible Solutions & Considerations

Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2

∙ Document oversight requirements for any AI agent deployments

Small Business (10-49 staff) / BLS Firm Size Classes 3-4

∙ AI agent governance policy with human oversight requirements

Medium Business (50-249 staff) / BLS Firm Size Classes 5-6

∙ Formal AI agent governance framework
∙ Human oversight requirements

Large Business (250-999 staff) / BLS Firm Size Classes 7-8

∙ AI agent governance program
∙ Formal oversight model with defined roles and controls

Enterprise (> 1,000 staff) / BLS Firm Size Class 9

∙ Enterprise AI agent governance framework
∙ Dedicated AI agent oversight team
∙ Zero-trust agent architecture

SCR-CMM

Level 0 Not Performed

Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed.

Level 1 Performed Informally

SCR-CMM Level 1 criteria definitions are not available for this control:
▪ A reasonable person would conclude this control requires a structured process.
▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality.

Level 2 Planned Tracked

Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist:
▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity.
▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners.
▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD).
▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines).
▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel.
▪ AAT is regarded as a technology and governed by the entity's existing IT governance practices.
▪ No formal Governance, Risk & Compliance (GRC) team exists to provide oversight of AAT-related activities. GRC functions are assigned to existing IT and/or cybersecurity personnel.

Level 3 Well Defined

Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist:
▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function.
▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners.
▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities.
▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform).
▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls).
▪ An implemented and operational capability exists to ensure AI agents are designed, developed and deployed to securely operate under human oversight.

Level 4 Quantitatively Controlled

Utilize SCR-CMM Level 3 criteria definitions:
▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control.
▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define.

Level 5 Continuously Improving

Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions:
▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control.
▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies.
▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.

1. Übersicht

Bezeichnung	Standard
Infrastructure Hardening & Isolation	Description Mechanisms exist to protect, isolate and harden infrastructure resources used by AI agents, including: (1) Resource allocation; (2) Privilege management; (3) Network segmentation; and (4) Workload isolation. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Use isolated environments for any AI agent deployments Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Network isolation and least-privilege for AI agent infrastructure Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal AI agent infrastructure hardening policy ∙ Isolation and privilege controls Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ AI agent infrastructure hardening program ∙ Network segmentation, privilege management Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise AI agent infrastructure security framework ∙ Zero-trust network architecture ∙ Container isolation (e.g., Kubernetes) SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to protect, isolate and harden infrastructure resources used by AI agents, including: (1) Resource allocation; (2) Privilege management; (3) Network segmentation; and (4) Workload isolation. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
AI Agent Limitations	Description Mechanisms exist to implement limitations for AI agents according to: (1) Least privileges, where the AI agent operates with the minimal permissions necessary to perform designated tasks; and (2) Least functionality, where the AI agent is restricted to communicate with the minimal Technology Assets, Applications and/or Services (TAAS) and networks necessary to perform designated tasks. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Apply least privilege to AI agent accounts/access Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Least-privilege and minimal footprint for AI agent configurations Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal AI agent access control policy with least-privilege requirements Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ AI agent privilege management program ∙ Automated enforcement of least-privilege Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise AI agent access governance platform ∙ PAM integration for AI agents ∙ Just-in-time access for AI agents SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to implement limitations for AI agents according to: (1) Least privileges, where the AI agent operates with the minimal permissions necessary to perform designated tasks; and (2) Least functionality, where the AI agent is restricted to communicate with the minimal Technology Assets, Applications and/or Services (TAAS) and networks necessary to perform designated tasks. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Tool & API Invocation Controls	Description Mechanisms exist to authenticate, authorize, validate and monitor all tool and Application Programming Interface (API) invocations by AI agents, including; (1) Schema validation; (2) Rate limiting; (3) Access controls; and (4) Output validation. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Authenticate and log AI agent tool/API calls Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ API key management and logging for AI agent tool calls Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal AI agent tool invocation control policy ∙ Authentication, authorization, and logging Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ AI agent API governance program ∙ Schema validation, rate limiting, and audit logging Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise AI agent API security platform ∙ Zero-trust API security ∙ Automated schema validation ∙ Rate limiting and anomaly detection SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to authenticate, authorize, validate and monitor all tool and Application Programming Interface (API) invocations by AI agents, including; (1) Schema validation; (2) Rate limiting; (3) Access controls; and (4) Output validation. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Orchestration Protocol Safeguards	Description Mechanisms exist to validate, secure and restrict AI agent orchestration protocols to prevent: (1) Unauthorized tool chaining; (2) Context manipulation; and/or (3) Protocol-based escalation. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Limit AI agents to approved tools only Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Whitelist approved tools/protocols for AI agents Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal AI agent orchestration security policy ∙ Protocol safeguards Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ AI agent orchestration security program ∙ Protocol validation and restriction controls Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise AI agent orchestration security framework ∙ Automated protocol enforcement ∙ Context integrity validation SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to validate, secure and restrict AI agent orchestration protocols to prevent: (1) Unauthorized tool chaining; (2) Context manipulation; and/or (3) Protocol-based escalation. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Data Pipeline & Input Integrity	Description Mechanisms exist to validate, sanitize and monitor all data inputs and retrieval pipelines for AI agents to: (1) Ensure data provenance; and (2) Prevent unauthorized access risks (e.g., injection, manipulation or exfiltration). Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Validate inputs to AI agents before processing Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Input validation and sanitization for AI agent data pipelines Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal AI agent input integrity policy ∙ Validation and sanitization procedures Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ AI agent data pipeline security program ∙ Automated input validation and provenance tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise AI agent data integrity platform ∙ Data provenance tools ∙ Automated sanitization pipelines SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to validate, sanitize and monitor all data inputs and retrieval pipelines for AI agents to: (1) Ensure data provenance; and (2) Prevent unauthorized access risks (e.g., injection, manipulation or exfiltration). Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Privileged Role & Delegation Boundaries	Description Mechanisms exist to prevent privilege escalation or unauthorized delegation by AI agents by: (1) Monitoring and enforcing dynamic roles; and (2) Establishing cross-agent delegation boundaries and privileged actions. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to prevent privilege escalation or unauthorized delegation by AI agents by: (1) Monitoring and enforcing dynamic roles; and (2) Establishing cross-agent delegation boundaries and privileged actions. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
AI Agent Data Access Restrictions	Description Mechanisms exist to restrict agent access to sensitive/regulated data so that AI agents cannot ingest, generate or act on unauthorized data. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to restrict agent access to sensitive/regulated data so that AI agents cannot ingest, generate or act on unauthorized data. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Data Extraction	Description Mechanisms exist to prevent AI agents from extracting sensitive/regulated data from volatile memory that can be exploited at a later point. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to prevent AI agents from extracting sensitive/regulated data from volatile memory that can be exploited at a later point. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
AI Agent Identity & Impersonation Defense	Description Mechanisms exist to ensure user identification and authentication methods are capable of preventing AI agents from (1) Spoofing; (2) Mimicry; and/or (3) Impersonation attacks. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to ensure user identification and authentication methods are capable of preventing AI agents from (1) Spoofing; (2) Mimicry; and/or (3) Impersonation attacks. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
AI Agent Logic Integrity	Description Mechanisms exist to prevent AI agent logic from being subverted or manipulated. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to prevent AI agent logic from being subverted or manipulated. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Sandboxing AI Agents	Description Mechanisms exist to utilize a "sandbox" capability to restrict AI agents from unrestricted access to both local and remote resources. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to utilize a "sandbox" capability to restrict AI agents from unrestricted access to both local and remote resources. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Prompt Injection Defense	Description Mechanisms exist to detect and mitigate prompt injection / input attacks that seek to manipulate AI agent instructions, bypass security, compliance and/or resilience controls or result in unauthorized actions. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to detect and mitigate prompt injection / input attacks that seek to manipulate AI agent instructions, bypass security, compliance and/or resilience controls or result in unauthorized actions. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Agent Kill Switch / User Control	Description Mechanisms exist to allow authorized users or operators to immediately halt or disable AI agent activity in case of unexpected behavior or harm. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to allow authorized users or operators to immediately halt or disable AI agent activity in case of unexpected behavior or harm. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Adversarial & Red Team Testing	Description Mechanisms exist to regularly conduct adversarial testing that simulates attacks against AI agents to identify and mitigate vulnerabilities. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to regularly conduct adversarial testing that simulates attacks against AI agents to identify and mitigate vulnerabilities. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Self-Modification Controls	Description Mechanisms exist to control, restrict and log AI agent self-modification. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to control, restrict and log AI agent self-modification. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Purging AI Agent Data	Description Mechanisms exist to restrict purging of any persistent memory or long-term data used by AI agents. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to restrict purging of any persistent memory or long-term data used by AI agents. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Delegation and Chaining Control	Description Mechanisms exist to restrict agentic delegation, chaining and multi-agent communication to prevent unauthorized task escalation or emergent risks. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to restrict agentic delegation, chaining and multi-agent communication to prevent unauthorized task escalation or emergent risks. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Behavioral Drift Detection	Description Mechanisms exist to continuously monitor for behavioral drift or deviation from established AI agent baselines. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to continuously monitor for behavioral drift or deviation from established AI agent baselines. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
AI Agent Action Authentication & Authorization	Description Mechanisms exist to ensure that all AI agent-initiated actions are properly mapped to authenticated user or system identities, with enforced authorization checks. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to ensure that all AI agent-initiated actions are properly mapped to authenticated user or system identities, with enforced authorization checks. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Transparency & Audit	Description Mechanisms exist to provide comprehensive audit trails of AI agent actions including: (1) Rationales; and (2) User/trigger mappings. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to provide comprehensive audit trails of AI agent actions including: (1) Rationales; and (2) User/trigger mappings. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Explainability	Description Mechanisms exist to: (1) Provide human-understandable explanations for significant AI agent actions or decisions; and (2) Enable users to contest outcomes. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to: (1) Provide human-understandable explanations for significant AI agent actions or decisions; and (2) Enable users to contest outcomes. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Ethics, Fairness & Bias Detection	Description Mechanisms exist to detect unfair, unethical or biased AI agent actions. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to detect unfair, unethical or biased AI agent actions. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.
Agent Output Integrity & Verification	Description Mechanisms exist to validate AI agent-generated outputs through the use of: (1) Content scanning; and (2) Output vetting through human approvals, where appropriate. Possible Solutions & Considerations Micro-Small Business (<10 staff) / BLS Firm Size Classes 1-2 ∙ Document relevant policy and procedures Small Business (10-49 staff) / BLS Firm Size Classes 3-4 ∙ Written policy and procedures ∙ Designated responsible owner ∙ Annual review Medium Business (50-249 staff) / BLS Firm Size Classes 5-6 ∙ Formal program with documented processes ∙ Regular review and testing Large Business (250-999 staff) / BLS Firm Size Classes 7-8 ∙ Enterprise program with dedicated resources ∙ Automated tooling ∙ Metrics tracking Enterprise (> 1,000 staff) / BLS Firm Size Class 9 ∙ Enterprise platform with dedicated team ∙ Automated monitoring ∙ Continuous improvement program SCR-CMM Level 0 Not Performed Practices are non-existent, based on the inability to demonstrate an implemented and operational capability. A reasonable person would conclude the control is not being performed. Level 1 Performed Informally SCR-CMM Level 1 criteria definitions are not available for this control: ▪ A reasonable person would conclude this control requires a structured process. ▪ At this level of maturity, the "ad hoc" nature of performing a capability informally would indicate the intent of the control is not met due to a lack of consistency and formality. Level 2 Planned Tracked Artificial Intelligence and Autonomous Technology (AAT) capabilities are requirements-driven, but are not standardized across the entity (e.g., local/regional level consistency). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are documented and maintained by process owners. ▪ IT and/or cybersecurity personnel work with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address applicable statutory, regulatory and/or contractual requirements for Technology Assets, Applications, Services and/or Data (TAASD). ▪ Artificial Intelligence (AI)-related controls are primarily administrative and preventative in nature (e.g., policies, standards, procedures & guidelines). ▪ Asset management may be a defined function (e.g., team or department) or assigned as an additional duty to existing IT and/or cybersecurity personnel. Level 3 Well Defined Artificial Intelligence and Autonomous Technology (AAT) capabilities are standardized across the entity for applicability to People, Processes, Technologies, Data and/or Facilities (PPTDF) to ensure consistency for Technology Assets, Applications, Services and/or Data (TAASD). Capability criteria associated with this control reasonably expect the following criteria to exist: ▪ Policies and standards associated with AAT domain capabilities are formally documented and centrally-managed by the entity's Governance, Risk & Compliance (GRC) team, or similar function. ▪ Standardized Operating Procedures (SOP) associated with AAT domain capabilities are well-documented and kept current by process owners. ▪ An Artificial Intelligence Governance (AIG) team, or similar function, is appropriately staffed and supported to implement and maintain AAT domain capabilities. ▪ Technology is leveraged to enhance the efficiency and accuracy of AI governance, risk management and compliance operations (e.g., dedicated AI governance platform). ▪ The entity's Governance, Risk & Compliance (GRC) team, or similar function, works with business stakeholders and process owners to appropriately scope and reasonably implement cybersecurity and data protection controls associated with AAT domain capabilities to address Minimum Compliance Requirements (MCR) (e.g., applicable statutory, regulatory and/or contractual requirements) and Discretionary Security Requirements (DSR) (e.g., entity-required controls). ▪ An implemented and operational capability exists to validate AI agent-generated outputs through the use of: (1) Content scanning; and (2) Output vetting through human approvals, where appropriate. Level 4 Quantitatively Controlled Utilize SCR-CMM Level 3 criteria definitions: ▪ There are no defined Level 4 criteria, since it is reasonable to assume a quantitatively-controlled process is not necessary to operationalize this control. ▪ While it may be possible to develop “metrics-driven” capabilities for this control, the criteria would be organization-specific to define. Level 5 Continuously Improving Utilize SCR-CMM Level 3 or Level 4 (if available) criteria definitions: ▪ There are no defined Level 5 criteria, since it is reasonable to assume a continuously-improving process is not necessary to operationalize this control. ▪ Level 5 capabilities should be considered “world-class” where the control builds on Level 4 capabilities, but are continuously improving through Artificial Intelligence (AI) and/or Machine Learning (ML) technologies. ▪ While it may be possible to develop responsive capabilities for this control through the use of AI and/or ML technologies, the criteria would be organization-specific to define.

1.1 Referenzen

1.2 Identifizierte Anforderungen

1.3 Related Regulations

2. Identifizierte Anforderungen

Anforderungen
Source	Anforderung

3. Related Regulations

Regulations
Source	Regulierung

Linked Issues

Issuelinks
Linktyp	Issue
is related to	Annual
is related to	relative Control Weighting = 05
is related to	Process
is related to	Protect
is related to	SCRM Focus Tier 2 OPERATIONAL
is related to	SCRM Focus Tier 3 TACTICAL
blocks	Inability to maintain individual accountability
blocks	Improper assignment of privileged functions
blocks	Privilege escalation
blocks	Unauthorized access
blocks	Lost, damaged or stolen asset(s)
blocks	Loss of integrity through unauthorized changes
blocks	Business interruption
blocks	Data loss / corruption
blocks	Reduction in productivity
blocks	Information loss / corruption or system compromise due to technical attack
blocks	Information loss / corruption or system compromise due to non‐technical attack
blocks	Loss of revenue
blocks	Cancelled contract
blocks	Diminished competitive advantage
blocks	Diminished reputation
blocks	Fines and judgements
blocks	Unmitigated vulnerabilities
blocks	System compromise
blocks	Inability to support business processes
blocks	Incorrect controls scoping
blocks	Lack of roles & responsibilities
blocks	Inadequate internal practices
blocks	Inadequate third-party practices
blocks	Lack of oversight of internal controls
blocks	Lack of oversight of third-party controls
blocks	Illegal content or abusive action
blocks	Inability to investigate / prosecute incidents
blocks	Improper response to incidents
blocks	Ineffective remediation actions
blocks	Expense associated with managing a loss event
blocks	Inability to maintain situational awareness
blocks	Third-party cybersecurity exposure
blocks	Third-party physical security exposure
blocks	Third-party supply chain relationships, visibility and controls
blocks	Third-party compliance / legal exposure
blocks	Use of product / service
blocks	Reliance on the third-party

Secure Controls Framework -
The Secure Controls Framework® (SCF)

"The SCF is the Common Controls Framework™ (CCF), the world's most comprehensive cybersecurity and data privacy metaframework - it is also free to use. The entire concept is building secure, compliant and resilient capabilities in the most efficient and cost-effective manner possible.

The SCF is more than just a unified control catalog, since its included content creates a playbook for Governance, Risk & Compliance (GRC) capabilities. Used globally by organizations of every size, the SCF is a robust and scalable solution for security, compliance and resilience controls. As a comprehensive security framework, the SCF maps 1,400+ controls across 200+ laws, regulations, and industry frameworks so you can implement once and comply everywhere.

Like it or not, cybersecurity is a protracted war on an asymmetric battlefield, where the threats are everywhere and as defenders we have to make the effort to work together to help improve cybersecurity and data privacy practices, since we all suffer when massive data breaches occur or when cyber attacks have physical impacts. Hackers share information on attack methods with other hackers, so why shouldn’t the good guys share information on how to best protect an organization? We decided to take action and make a difference, since we feel it is too important to wait for someone else to fix the problems that exist.

The SCF is made up of volunteers, mainly specialists within the cybersecurity profession, who focus on GRC and the cybersecurity side of data privacy. These are auditors, engineers, architects, incident responders, consultants and other specialists who live and breathe these topics on a daily basis. The end product is "expert-derived content" that makes up the SCF." https://securecontrolsframework.com/

Terms & Conditions

The SCF End User License Agreement (EULA) governs the use of the Secure Controls Framework® (SCF) under the Creative Commons Attribution-No Derivatives 4.0 International Public License.