Do AI projects still need to meet the government Service Standard?
Answered
Yes. AI services must meet the same standards as other technology. If you build a service, meet the Service Standard (TCoP point 13). Point 9 requires Secure by Design principles; point 11 expects teams to understand AI technologies and avoid tech choices that undermine reliable information or decisions. All government AI must be secure and resilient per security.gov.uk. Follow Secure by Design / Service Manual security expectations.
From the guidance
Primary (how) Service Manual: Using artificial intelligence (AI) in services
Section: Using artificial intelligence (AI) in services
Read this in Service Manual: Using artificial intelligence (AI) in services (opens in new tab)
Secondary (normative) Service Standard: 9. Create a secure service which protects users’ privacy
Section: 9. Create a secure service which protects users’ privacy
Secondary (normative) Service Standard: 11. Choose the right tools and technology
Section: 11. Choose the right tools and technology
Read this in Service Standard: 11. Choose the right tools and technology (opens in new tab)
Secondary (normative) Technology Code of Practice
Section: 13. Meet the Service Standard
Secondary (normative) AI Playbook for the UK Government
Section: Principle 5: You understand how to manage the full AI life cycle
Read this in AI Playbook for the UK Government (opens in new tab)
Secondary (normative) About Secure by Design
Section: Secure by Design’s wider context
Secondary (normative) Artificial Intelligence (security.gov.uk)
Section: Using AI securely
Read this in Artificial Intelligence (security.gov.uk) (opens in new tab)
Related questions
- How does the Service Standard affect whether I should use AI?
- What does the Service Standard say about choosing AI technology?
- Which organisations must use the Algorithmic Transparency Recording Standard? Lawful, ethical and responsible use
- Do I need to use the Algorithmic Transparency Recording Standard (ATRS)? Lawful, ethical and responsible use
- Who should own cyber security risk for an AI service under Secure by Design?