AI Playbook for the UK Government
A 117-page guide from the UK’s Government Digital Service on adopting AI across government departments: 10 top-level principles, understanding AI, building AI solutions, buying AI, legal considerations, data protection, security, and governance, plus an appendix of public-sector case studies.
License: Open Government Licence v3.0 — © Crown copyright 2025.
Key points
- A chatbot response should carry a plain-text disclosure, not just a styled visual marker — the playbook’s own example: “this response has been written by an automated AI chatbot.” Useful wherever AI output travels somewhere a styled UI marker won’t (copied text, plain-text email, an API response). See AI Content Marking.
- When a system can’t fit real-time human review into the response loop (chatbots, other instant-response AI), human control has to move to other stages: pre-deployment testing, ongoing assurance checks on the live system, and a visible user-facing way to report a bad response that triggers review after the fact. See Human-AI Interaction.
- UK GDPR Article 22 prohibits any decision based solely on automated processing that has a legal or similarly significant effect on a person, unless a human is meaningfully involved — a harder-edged rule than a mere appeal channel once a decision crosses that stakes threshold. See Human-AI Interaction.
- A visible AI-content marker doesn’t survive content leaving the interface that showed it (a screenshot, a copy-paste, a redistributed image or video) — pairing it with a machine-readable signal embedded in the content itself, such as a watermark or C2PA provenance metadata, lets the AI origin still be verified downstream. See AI Content Marking.