With a $1 million investment and the backing of Meta and the Alan Turing Institute, the UK government is launching a program aimed at attracting the country’s top AI engineers to develop real solutions in urban planning, national security, and citizen services.
The UK’s Department for Science, Innovation and Technology (DSIT) has announced the launch of the Open-Source AI Fellowship, a pioneering initiative set to reshape the relationship between artificial intelligence and public administration. Funded with $1 million from Meta and managed by the Alan Turing Institute, this fellowship will allow selected AI engineers to work for 12 months within the UK government, creating practical tools using open-source models like Llama 3.5.
This isn’t about academic or experimental work; the goal is to solve genuine state problems using transparent, reusable, and accessible AI technologies.
### From Theory to Practice: Applying AI to Public Challenges
During their tenure, fellows will participate in high-impact projects where AI can produce tangible change. Some already-defined examples include:
– Secure translation of confidential documents in intelligence and diplomacy contexts, utilizing locally-run AI in controlled environments.
– Automating urban planning processes by leveraging policy and regulation data to speed up construction permits.
– Offline backup systems for the NHS and emergency services, useful during network outages or blackouts.
– Administrative decision-making assistants trained on legislation and public guidelines to improve process efficiency.
All solutions will be developed with open-source models and published as free software to promote reuse by other governments, businesses, and institutions.
### Caddy and Humphrey: Functional AI in the UK Government
This announcement follows the success of two previous initiatives already showing promising results:
– Caddy, an AI assistant developed in collaboration with Citizens Advice, is now used in six call centers and the Cabinet Office. It helps operators find relevant information in real-time during calls about benefits, consumer rights, or debts. Initial tests showed it reduced average response time by half, with 80% of generated responses used without modification.
– Humphrey comprises tools that assist officials with repetitive tasks such as summarizing documents, drafting reports, or compiling responses to public inquiries. It’s used internally to free up staff time from bureaucratic workloads.
Both projects reinforce the UK government’s approach: AI as a practical infrastructure to enhance public services, not just a showcase or theoretical solution.
### Building Sovereign Capabilities with Open AI
Choosing open-source models like Llama 3.5, supported by Meta, is deliberate. The government wants to avoid dependence on proprietary solutions from major tech companies and reduce the costs of large-scale deployment. These models also give public agencies control over data, parameters, and local execution—crucial in areas like national security or healthcare.
“Our aim is that no official spends time on tasks that AI can do better and more cheaply,” recently stated the British prime minister. Meanwhile, Tech Secretary Peter Kyle emphasized that this fellowship represents an “action-oriented, citizen-service-driven AI based on collaboration and open code.”
### Shared Resources Platform and a Collaborative Culture
Alongside the fellowship launch, a new phase of the AI Knowledge Hub has been introduced (https://ai.gov.uk/knowledge-hub), a platform that gathers real-world examples, tools, prompts, best practices, and success stories to facilitate AI adoption across all UK government departments.
The Hub aims to avoid duplicate efforts, scale proven solutions, and foster a shared, responsible AI culture within the UK’s public ecosystem.
—
### Key Aspects of the Open-Source AI Fellowship Program
| Aspect | Details |
|———————-|————————————————————–|
| Duration | 12 months (starting January 2026) |
| Funding | $1 million (Meta → Alan Turing Institute) |
| Participants | AI experts based in the UK |
| Technologies | Open-source models like Llama 3.5 |
| Objectives | National security, urban planning, offline AI |
| Environment | Direct integration into UK government teams |
| Expected Outcomes | Functional tools published as free software |
—
### AI as State Infrastructure
With this initiative, the UK advances its commitment to turning artificial intelligence into a strategic asset for the public sector—not merely as an add-on, but as an integral part of operational infrastructure.
The combination of domestic talent, institutional support, open technologies, and focus on real-world problems reinforces a pragmatic vision of AI: tools that are used, adapted, and shared, with a direct impact on citizens’ lives.
The call for applications is now open, and competition among candidates is expected to be fierce. Official submissions will begin soon, marking the start of a new phase in AI applied to public services.