Every AI Policy Ever · open in the interactive register
Published (non-binding, current)Non-binding · Voluntary code / commitments
Code of Practice for the Cyber Security of AI
Record GB-06
- Issued by
- Department for Science, Innovation and Technology (DSIT) (with DCMS)
- Jurisdiction
- United Kingdom
- Level · region
- National · Europe
- Published / adopted
- 2025-01-31
- Language
- English
Summary
A voluntary Code of Practice setting baseline cyber security requirements for AI systems, including generative AI, against AI-specific risks such as data poisoning, model obfuscation and indirect prompt injection. Its 13 principles span five lifecycle phases, from secure design to secure end of life, and assign provisions to developers, system operators and data custodians, among them capabilities for human oversight, security testing before release and deployment, monitoring of system behaviour, and incident management and recovery plans. It is meant as the basis of a global ETSI standard.
Safety topics
evaluation & testinghuman oversightaccountability & liabilitytransparency & labelling
Open question
the checker found something possibly newer: https://www.ncsc.gov.uk/blog-post/new-etsi-standard-protects-ai-systems-from-evolving-cyber-threats Issue #122
The record shows its last checked values until a maintainer decides; the decision is recorded in its history.
Status notes
Published with an implementation guide that DSIT commissioned and that DSIT and NCSC officials reviewed. The Code is an addendum to the Software Code of Practice and builds on the NCSC's Guidelines for Secure AI Development of November 2023; 80% of respondents to DSIT's call for views, held from 15 May to 9 August 2024, endorsed the intervention. DSIT, with the NCSC, will submit the Code and the guide to ETSI as the basis for a new global standard (TS 104 223) and implementation guide (TR 104 128), and the government will update the Code to mirror the ETSI standard.
Local copy
Archived. Last attempt 2026-10-17 20:42:09 UTC
Official source (third-party copy, not published) archive/GB-06/2026-10-11/official.html
Retrieved 2026-10-11 14:03:19 UTC
SHA-256 40888dd912eb090ddf65de325e278ab87bea4680afd406a59b49a19c976a06f3
Secondary source (third-party copy, not published) archive/GB-06/2026-10-11/secondary.html
Retrieved 2026-10-11 14:03:24 UTC
SHA-256 ddab8ffb20dee5fb183e79cb83fdeaddef841bdc75c934e1c1729fd66f62bd2e
Verification
AThe instrument's own official text, or an official register, was stored and confirms the title, date and status. checked 2026-10-17 · official-url-fetched
Stored and read in this run: the Code's GOV.UK HTML text (published 31 January 2025; scope, audience, the 13 principles and their provisions) and the GOV.UK publication page (stored as the secondary source; DCMS and DSIT, published 31 January 2025 with no later update, the implementation guide and the ETSI plan). Snapshots of 11 Oct 2026, unchanged on 17 Oct. Whether ETSI has since published TS 104 223, and whether the Code was updated to mirror it, is not in a stored file; the publication page showed no update on 11 Oct 2026.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-17: 5 field(s) supported.
History
- 2026-10-17notes: Published with an Implementation Guide. The government said it would submit the → Published with an implementation guide that DSIT commissioned and that DSIT and — the sentence 'Whether the ETSI standard has since been published was not verified here' is research language and moved to verification.note; notes extended with the guide, the Software Code addendum, the NCSC guidelines, the call for views and the ETSI numbers2026-10-17-recheck-europe-2 · Claude Code (claude-opus-5-5)
- 2026-10-17safety_topics: ["critical infrastructure", "accountability & liability"] → ["evaluation & testing", "human oversight", "accountability & liability", "trans — 'critical infrastructure' dropped: the Code covers AI systems in any sector, and 'Secure your infrastructure' (Principle 6) means an organisation's own infrastructure (the 2026-10-11 quality check flagged this); 'evaluation & testing' (Principle 9), 'human oversight' (Principle 4) and 'transparency & labelling' (Principles 8 and 10) added; 'accountability & liability' kept (Principle 4, audit trails)2026-10-17-recheck-europe-2 · Claude Code (claude-opus-5-5)
- 2026-10-17summary: A voluntary baseline of cyber security requirements for AI systems, including ge → A voluntary Code of Practice setting baseline cyber security requirements for AI — extended with the Code's own content (AI-specific risks, the stakeholder roles, human oversight, testing, monitoring, incident and recovery plans), so that the topics rest on the summary2026-10-17-recheck-europe-2 · Claude Code (claude-opus-5-5)
Citation
Department for Science, Innovation and Technology (DSIT) (with DCMS). Code of Practice for the Cyber Security of AI. 2025-01-31. https://www.gov.uk/government/publications/ai-cyber-security-code-of-practice/code-of-practice-for-the-cyber-security-of-ai. Status: Published (non-binding, current); verified 2026-10-17 (level A). Local copy archive/GB-06/2026-10-11/official.html retrieved 2026-10-11 14:03:19 UTC, SHA-256 40888dd912eb090ddf65de325e278ab87bea4680afd406a59b49a19c976a06f3. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record GB-06.
Improve this record
Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.