The UK government has announced a 12-month initiative designed to embed top-tier AI engineers from other organisations within Whitehall departments to build deployable, open-source tools for public service use.
Funded by a $1m grant from Meta and delivered in partnership with the Alan Turing Institute, the new Open-Source AI Fellowship aims to create transparent, sovereign AI infrastructure that can scale across departments while reducing costs and enhancing productivity.
Positioned as a practical response to growing demand for generative AI talent in both public and private sectors, the fellowship will provide a route for engineers to work on high-impact public sector problems — from accelerating planning approvals using structured data to developing secure offline AI systems for emergency services.
Fellows will work with open-source foundation models such as Meta’s Llama 3.5 and will join the Department for Science, Innovation and Technology’s (DSIT) Incubator for AI, the team behind Humphrey — a digital assistant currently used in government to summarise documents, draft briefings and reduce administrative overhead.
The fellowship forms part of the government’s broader AI modernisation drive. Recent examples include Caddy, a customer service assistant developed with Citizens Advice, now being used by the Cabinet Office to surface grant decision guidance in real time.
Internal trials showed Caddy halved response times, with 80% of AI outputs requiring no edits and advisors reporting significantly increased confidence in their responses.
UK Technology Secretary Peter Kyle said the fellowship was designed to “scale real, working tools across government” and build “sovereign capabilities where the UK must lead, like national security and critical infrastructure.”
Meta’s chief global affairs officer Joel Kaplan said the initiative underscored the value of open-source models in the public domain: “This partnership with the Alan Turing Institute will help the government access some of the brightest minds and the technology they need to solve big challenges – and to do it openly and in the public interest.”
Dr Jean Innes, CEO, Alan Turing Institute, said the programme would offer “an innovative way to match AI experts with the real-world challenges our public services are facing.”
Applications for the Open-Source AI Fellowship will open next week via the Alan Turing Institute, with the first cohort beginning in January 2026.
Use cases will be selected in collaboration with departments and all tools will be open sourced for reuse.
The government has also launched a new phase of its internal AI Knowledge Hub, adding new features such as a prompt library to accelerate practical adoption and reduce duplication across teams.
Similar initiatives are gaining traction in the US, where the US Digital Corps and Presidential Innovation Fellows programmes embed technologists into federal agencies to support mission-driven innovation.
These initiatives actively recruit from the private sector and prioritise open innovation. In addition, the US has launched the National AI Research Resource (NAIRR) pilot — a public-private initiative involving industry leaders and federal agencies to expand access to compute, data, and tools for responsible AI development in the public interest.
