10 principles for using AI

These principles apply to using Artificial Intelligence (AI) in government, and are available as AI Playbook posters on the GOV.UK website.

Understand AI and its limits

Understand the technology you’re using, what it can and cannot do, and the potential risks it poses. AI outputs are not always accurate, you must verify them and manage potential harm.

Use AI lawfully, ethically and responsibly

Use AI lawfully and responsibly. Get early advice on legal issues, such as, data protection, intellectual property, and equality. Design your project to address ethical concerns, manage bias, and have a positive impact on the public and environment.

Know how to use AI securely

AI products must be secure and resilient to cyber attacks. Comply with government security standards and the Secure by Design principles. Understand risks specific to AI (for example, data poisoning and prompt injections). Mitigate these by implementing safeguards, such as, security testing and content filtering.

Have meaningful human control at the right stages

Humans must have meaningful control over the AI system throughout the product’s lifecycle. Continuously monitor its behaviour and have plans to prevent harm. Humans must validate all high-risk decisions influenced by AI, and users must be able to report issues for human review.

Understand how to manage the AI life cycle

Plan to manage the AI solution across its life cycle, for example, set-up, maintenance, updates, and retirement. You must also monitor and mitigate model drift, bias, and hallucinations.

Use the right tool for the job

Choose the most appropriate technology for your needs. Be open to AI solutions, but always consider if the problem could be solved more easily with traditional or simpler technologies.

Be open and collaborative

Work with teams across government to share expertise and avoid duplicating effort. Be open with the public about how and where AI systems are being used, for example:

Work with commercial colleagues from the start

Contact commercial colleagues early to get specific advice on the rapidly developing AI market. Use contracts to make sure external suppliers meet the same ethical standards and are transparent about their AI systems.

Have the skills and expertise you need

Your team must have the technical and ethical skills to implement and use AI safely. Decision makers and senior leaders should understand the risks and opportunities of AI.

Use these principles alongside existing policies

Use these principles alongside your organisation’s existing governance structures and policies, especially for security and data handling. Consider the implications of AI when drafting new policies. Connect with assurance teams early in the project life cycle and have clear review and escalation processes in place.

Back to top