Government launches $1 million fellowship to supercharge public sector AI

The UK government has unveiled a new $1 million initiative to attract top-tier AI talent into the heart of the public sector. Funded by a grant from Meta and delivered through the Alan Turing Institute, the 12-month Open-Source AI Fellowship aims to build cutting-edge, open-source tools that directly support public service delivery, improve national security capabilities, and boost productivity across government departments.

Announced by the Department for Science, Innovation and Technology (DSIT) and Technology Secretary Peter Kyle, the programme invites the UK’s best AI engineers to undertake a “tour of duty” in government. These fellows will work on high-impact, real-world challenges using open-source AI models such as Meta’s Llama 3.5, with a strong focus on transparency, cost-efficiency, and public good.

“This Fellowship is the best of AI in action – open, practical, and built for public good,” Kyle said. “It’s about delivery, not just ideas – creating real tools that help government work better for people.”

AI in action: Caddy and Humphrey

The launch comes as the government begins rolling out “Caddy”, an AI-powered customer service assistant developed in partnership with Citizens Advice. Caddy supports frontline call handlers by surfacing expert guidance in real time. Currently used across six Citizens Advice call centres, early tests showed Caddy halved response times and doubled staff confidence in providing accurate answers. The tool has now been deployed within the Cabinet Office to assist with grant decision-making.

Alongside Caddy, the government has also been trialling “Humphrey” - a suite of AI tools that helps Civil Servants process ministerial requests by automating tasks like summarising documents and drafting consultation responses. Fellows will have the opportunity to enhance and expand these systems, making them available to more departments and use cases.

National security and sovereign AI capabilities

Some fellows will be embedded within departments working on sensitive use cases. These include AI-powered translation tools for national security operations, as well as secure assistants capable of operating within closed, offline systems used for emergencies or network failures.

The use of open-source models is central to the programme’s design. By relying on publicly available technologies, the government hopes to reduce long-term AI costs and avoid dependence on proprietary systems. DSIT estimates that open-source AI could unlock up to £45 billion in productivity gains across the public sector.

Meta’s $1 million grant to the Alan Turing Institute provides the backbone of the fellowship, with fellows set to join DSIT’s Incubator for AI - the innovation unit behind both Caddy and Humphrey. Dr Jean Innes, CEO of the Alan Turing Institute, added: “These fellowships will offer an innovative way to match AI experts with the real-world challenges our public services are facing.”

Scaling AI across government

The fellowship is part of a broader push to integrate AI more deeply and responsibly into government operations. Alongside the announcement, the government has also launched the next phase of its AI Knowledge Hub, a platform to help departments share best practice and scale successful AI pilots. The Hub will soon include a Prompt Library - curated examples and templates to help civil servants make everyday use of AI tools.

Applications for the Open-Source AI Fellowship will open next week, with placements beginning in January 2026. More information and registering of interest can be found here