Artificial intelligence (AI) has the potential to help councils improve services and work more efficiently. To realise those benefits responsibly, councils need clear governance, which is usually set by an organisation's AI policy.
As part of the Local AI team's work to support safe and responsible AI adoption, we wanted to better understand how councils are developing and implementing AI policies. By identifying common challenges, gaps in available support, and examples of emerging practice, we can develop shared resources that provide a consistent baseline and avoid unnecessary duplication of effort.
What we did
The Responsible AI team within Local AI interviewed local government officials from over 15 councils who had been involved in developing AI policies within their organisations. The discussions explored how councils approached policy development, the impacts of those policies and the challenges they encountered along the way.
We heard from councils at very different stages of the journey. Some were developing their first policy, while others were already revisiting theirs, as technology, organisational understanding and the ways staff use AI evolved.
What came through strongly was an enthusiasm for sharing what had been learned. There were plenty of questions still being worked through, but also a real willingness to learn from each other and help others avoid having to start in the same place.
What we learned
Develop the policy with people across the council
AI policy is rarely the work of one team. Councils told us about bringing together colleagues from information governance, digital and IT, analytics, HR, finance, legal services, content design, trade unions and other parts of the organisation. That breadth matters. Different teams spot different risks and opportunities, and involving them early can help build wider understanding, ownership and support for the policy. Devon County Council, for example, set up an AI Steering Group and AI Board with representatives from across the organisation to oversee AI-related activity and consider its ethical implications.
Integrate AI policies into a wider governance system, so the policy doesn’t govern on its own
Councils we spoke to are supporting their policies with practical mechanisms that help staff know what they can use and what to do when a new opportunity arises. Denbighshire County Council, for example, has integrated its AI policy into its AI request process, helping ensure that people requesting access to AI tools are aware of the policy and its requirements. Another council we spoke to uses an AI Ethics Board to review proposed uses of AI across services. Another has introduced new technical controls to restrict access to unapproved AI services. These mechanisms turn written policies into operational decisions. Without them, staff may understand the policy but still be unsure what to do with it.
Don’t stop at just publishing the AI policy
Putting an AI policy on the intranet does not automatically mean people will know about it, understand it or know how it applies to their work. Councils we spoke to used combinations of staff bulletins, manager briefings, workshops, videos, dedicated AI pages, learning pathways and AI champions to bring their policies to life. Leeds City Council, for example, has incorporated AI into mandatory information governance training, developed a dedicated intranet area with guidance and training, and created an in-house 45-minute AI training course alongside further learning and supporting materials. Approaches like this help turn the policy from a static document into something that supports day-to-day decisions about responsible AI use.
Think about transparency from the start
Councils are also considering when and how residents should know that AI has contributed to a service, communication or decision. There is unlikely to be one answer for every use of AI. The level of transparency needed for a low-risk use may look very different from something that could significantly affect a resident. Councils are thinking about how people should be told about the use of AI in public services, what human oversight is in place, and what alternative routes to a service are available. Building these questions into the design of services is much easier than trying to add transparency after an AI tool has already been introduced.
Making AI policy easier to develop and use
Many councils have learned from their peers, adapting policies developed elsewhere and drawing on existing information management, data governance and other policies. While these are useful resources, a clear message from our interviews was the need for more practical support. Several officials said they would have found it helpful to have access to a template or framework covering the key issues an AI policy should address, while still allowing flexibility to reflect local governance structures, terminology and risk appetite.
Some councils also noted that existing guidance could be lengthy or difficult to navigate, making it harder to translate high-level principles into something practical. This suggests a need for resources that are both accessible and actionable, helping councils move from understanding responsible AI principles to implementing them.
This is where we hope our upcoming work can help.
What's next?
The findings of these interviews alongside broader engagement with local government stakeholders have been shaping the Responsible AI team's work to support councils adopting AI responsibly.
AI policy template
We are developing an AI policy template for councils, drawing on feedback from local authorities and a review of more than 40 council AI policies. It will provide a practical starting point for councils developing a policy for the first time, as well as those reviewing an existing one. The template helps councils ask the right questions, define responsibilities and support responsible adoption.
Responsible AI Charter
More immediately, the team is co-designing a Responsible AI Charter with councils. The Charter will set out a shared baseline of commitments for responsible AI adoption, providing something councils can build from while retaining the flexibility to decide how those commitments are put into practice locally.
We have already run several workshops to help co-design these products which have led to significant improvements. We are hoping to publish these resources by spring 2027.
If you'd like to be a co-design partner and help us shape the AI policy template for local government, please don't hesitate to get in touch with us by emailing LocalAI@communities.gov.uk.
Leave a comment