Code of Practice for the Cyber Security of AI

• Authority: Government of the United Kingdom
• Jurisdiction: Other countries
• Collections: Miscellaneous documents
• Most recent activity: Proposed
• Most recent activity date: 2025-01-31

Risk subdomain coverage

Good coverage — Comprehensive governance measures or mitigations with clear procedures relating to the risk subdomain are explicitly described.

• Compromise of privacy
• AI system security vulnerabilities
• Cyberattacks, weapon development or use
• Lack of capability or robustness
• Lack of transparency or interpretability

Minimal coverage — Brief mention but no specific mitigations, recommendations or governance measures relating to the risk subdomain are described.

• Overreliance and unsafe use
• Loss of human agency and autonomy
• Competitive dynamics
• Governance failure
• Multi-agent risks

All other risk subdomains had no coverage in this document.

Risk Coverage Summary

The document has strong coverage of approximately 8-10 subdomains, with primary focus on AI system security (2.2), malicious actors and cyberattacks (4.2), AI safety failures including lack of robustness (7.3), lack of transparency (7.4), and multi-agent risks (7.6). It also addresses governance failure (6.5) through security governance frameworks, and competitive dynamics (6.4) implicitly through secure development practices.

Sector coverage

Good coverage — Comprehensive governance measures or mitigations with clear procedures relating to the sector are explicitly described.

• Information
• Scientific Research & Development Services

All other sectors had no coverage in this document.

Sectors Summary

This is a cross-sectoral Code of Practice that applies broadly to all organizations developing, deploying, or operating AI systems regardless of industry. The document does not specify particular economic sectors but provides universal security guidance applicable across all sectors where AI is used.

Legislative Status

Soft Law

Legislative Status Summary

This is a voluntary Code of Practice with predominantly mandatory language ('shall') but no formal enforcement mechanisms, penalties, or sanctions. It relies on voluntary adherence and industry best practices.

Actors

Proposer

• Entity Types: AI_Governance_Actor
• Entities: Government of the United Kingdom
• Summary: The document is explicitly authored by the Government of the United Kingdom as stated in the document information header.

Target

• Entity Types: AI_Developer, AI_Deployer
• Entities: Developers; System Operators; Data Custodians; End-users
• Summary: The document explicitly targets multiple actor types throughout all principles. Developers are those who build AI systems, System Operators are those who deploy and operate them, Data Custodians manage data, and End-users are those who use the systems. The document applies primarily to Developers and System Operators as indicated in each principle.

Enforcer

• Summary: No enforcement body, agency, or authority is specified in the document. As a Code of Practice, it appears to rely on voluntary compliance without formal enforcement mechanisms.

Monitor

• Summary: No specific monitoring body or oversight agency is identified in the document. While the document requires organizations to monitor their own systems, no external monitoring authority is designated.

AI Lifecycle Stages

Lifecycle stages covered by the documents, as defined in the OECD Framework for the Classification of AI Systems

• Plan and Design
• Collect and Process Data
• Build and Use Model
• Verify and Validate
• Deploy
• Operate and Monitor

AI Lifecycle Summary

The document comprehensively covers all stages of the AI lifecycle from planning and design through end-of-life disposal. It emphasizes secure design (Principles 1-4), secure development (Principles 5-9), secure deployment (Principle 10), secure maintenance (Principles 11-12), and secure end-of-life (Principle 13).

Technical Scope

Terms relating to technical scope that are mentioned or implicitly referenced in the document

• AI Models
• AI Systems

Technical Scope Summary

The document extensively covers AI systems and AI models throughout. It does not explicitly define or mention frontier AI, general purpose AI, task-specific AI, foundation models, generative AI, predictive AI, open-weight models, or specific compute thresholds. The focus is on AI systems broadly defined with security considerations.