The UK government has published a new artificial intelligence assurance framework requiring public sector bodies to test, document and monitor AI systems before they are deployed in live services. The guidance, issued by the Department for Science, Innovation and Technology, sets out the first common national standard for the safe use of AI across central and local government.
The framework introduces mandatory risk assessments for high-impact systems, ongoing performance monitoring, and a requirement for officials to publish summary records of how AI tools are used. It applies to everything from chatbots on council websites to automated decision-making in benefit and licensing processes, and follows a year of pilot testing with 22 public bodies.
Government figures show that more than 500 AI use cases are now live across the public sector, with spending on AI-related contracts exceeding £340 million in the last financial year. Officials said the framework was designed to build public trust rather than slow adoption, arguing that poorly governed projects cost more in the long run.
“This is about making sure every deployment is safe, transparent and accountable from day one,” a spokesperson for the Department for Science, Innovation and Technology said. “The public sector is already a major user of AI, and these standards will help it use the technology well.”
Digital policy analysts broadly welcomed the move, though several noted that enforcement remains voluntary for local authorities. One former government technology adviser said the framework’s success would depend on whether the bodies that adopt it are properly resourced to carry out the checks it requires.
The framework also aligns with the UK’s wider approach to AI regulation, which relies on existing regulators rather than a single new law. Ministers have resisted calls for a dedicated AI bill, arguing that sector-specific rules are more flexible as the technology evolves.
A consultation on the assurance framework will run until late autumn, with final guidance expected early next year. Officials confirmed that a central register of public sector AI deployments would follow, giving citizens a single place to see which systems are in use and how they are governed.