How should I secure the AI supply chain (models and components)?

Answered

Assess and monitor AI supply-chain security across the life cycle. Require suppliers to meet the same standards you apply to other software; if they cannot, follow your risk policy. Treat third-party models and serialised weights as untrusted code — scan and sandbox on import. Secure by Design also requires continual due diligence on third-party platforms, software and code.

From the guidance

Primary (how) NCSC: Secure development (AI systems)

You assess and monitor the security of your AI supply chains across a system’s life cycle, and require suppliers to adhere to the same standards your own organisation applies to other software.

Section: Secure your supply chain

Read this in NCSC: Secure development (AI systems) (opens in new tab)

Secondary (normative) NCSC: Secure design (AI systems)

you implement scanning and isolation/sandboxing when importing third-party models or serialised weights, which should be treated as untrusted third-party code and could enable remote code execution

Section: Design your system for security as well as functionality and performance

Read this in NCSC: Secure design (AI systems) (opens in new tab)

Secondary (normative) Secure by Design Principles

Where third-party products are used, perform security due diligence by continually assessing platforms, software and code for security vulnerabilities.

Section: 2. Source secure technology products

Read this in Secure by Design Principles (opens in new tab)

Secondary (normative) Code of Practice for the Cyber Security of AI

Developers and System Operators shall follow secure software supply chain processes for their AI model and system development.

Section: Principle 7: Secure your supply chain

Read this in Code of Practice for the Cyber Security of AI (opens in new tab)

Related questions