The British government has unveiled a multi-pronged artificial intelligence initiative designed to modernize essential public services through strategic partnerships with Meta and Anthropic, placing leading AI researchers inside Whitehall to develop practical tools for transport maintenance, employment support, and national security operations.
Open Source Fellowship Brings Private Sector Expertise Into Government
The Department for Science, Innovation and Technology (DSIT) is embedding a cohort of AI specialists within its operations through a $1 million Open Source AI Fellowship administered by the Alan Turing Institute. The 12-month program, funded by Meta and operational from January 2026, tasks fellows with building tools that remain government-owned and publicly accessible.
Fellows will operate within DSIT’s Incubator for Artificial Intelligence, the same unit responsible for developing Humphrey, a suite of AI tools already in use across government for document summarization, note-taking, and consultation analysis. The initiative reflects a deliberate strategy to reduce dependency on proprietary commercial systems by leveraging open-source models including Meta’s Llama family.
Computer Vision for Infrastructure and Secure Systems for Defense
The program’s technical focus spans both civilian infrastructure and sensitive national security applications. In transport, fellows will develop computer vision models capable of analyzing images and video feeds to help local councils prioritize road maintenance and infrastructure repairs, a capability that could fundamentally change how authorities identify and address deteriorating assets across thousands of miles of roads and bridges.
For defense and intelligence operations, the fellowship will produce AI systems engineered to function entirely within air-gapped networks. These secure environments require specialized models capable of processing classified documents without external connectivity, addressing a persistent challenge in deploying advanced AI where data exposure risks are unacceptable. Translation tools for national security contexts represent one immediate application.
The program also targets emergency scenarios where network infrastructure fails. Fellows are expected to develop offline-capable AI systems that can support NHS staff and emergency responders during power outages or communication blackouts—situations where AI assistance may be most valuable but traditional cloud-based systems become inaccessible.
Anthropic Partnership Targets Employment Services Through GOV.UK
Separately, Anthropic has secured a contract to build an AI-powered assistant for GOV.UK, the central portal for government services. The system will initially focus on employment support, guiding job seekers through available training programs, explaining eligibility for benefits, and providing personalized career advice based on individual circumstances.
Unlike conventional chatbots, the planned assistant represents an agentic system designed to maintain context across multiple interactions. Users returning to the service will not need to re-enter information from previous sessions, and the system will actively route people toward appropriate services rather than simply answering questions. Anthropic engineers will work directly alongside developers at the Government Digital Service throughout the project, with an explicit goal of building internal capacity so the UK government can maintain the system independently.
The partnership formalizes a February 2025 Memorandum of Understanding between DSIT and Anthropic’s CEO, and represents one of the first tangible outcomes from that agreement. A pilot is expected later this year, following DSIT’s “Scan, Pilot, Scale” framework that tests new technologies before broader deployment.
Caddy Demonstrates Open Source Potential in Call Centers
The fellowship launch coincides with the open-sourcing of Caddy, an AI assistant developed in partnership with Citizens Advice that helps call center staff access guidance documents rapidly. Currently deployed across six Citizens Advice locations, the tool assists advisors handling queries on debt management, consumer rights, and legal referrals.
Initial testing across approximately 1,000 calls indicated response times could be halved, with 80% of AI-generated responses requiring no revision before use. Advisors using the system reported twice the confidence in providing accurate answers. The Cabinet Office has now adopted Caddy to help teams access expert guidance on grant-making decisions—its first deployment within central government.
By releasing Caddy’s code publicly, DSIT is positioning the tool for adoption by call centers worldwide, demonstrating the government’s commitment to generating broader public benefit from taxpayer-funded AI development.
CustomerFirst Unit Anchors Broader Digital Transformation
The AI initiatives operate within a larger digital transformation agenda anchored by CustomerFirst, a new DSIT unit established in January 2026. Led by Tristan Thomas, formerly of digital bank Monzo, and co-chaired by Octopus Energy CEO Greg Jackson, the unit brings private sector expertise into public service redesign.
The Driver and Vehicle Licensing Agency serves as the initial testing ground, with the unit tasked with rethinking how DVLA handles millions of annual interactions involving driving licenses and vehicle registration. DSIT estimates the approach could enable £4 billion in taxpayer savings by shifting service fulfillment from phone and postal channels to online platforms.
Octopus Energy’s experience provides a benchmark: the company uses generative AI tools to assist in drafting 35% of customer emails while maintaining 70% customer satisfaction ratings. CustomerFirst aims to replicate such efficiency gains across government services while preserving access for those less confident with technology.
£2 Billion AI Investment Underpins Long-Term Strategy
The fellowship and employment assistant projects represent components of the UK’s broader AI Opportunities Action Plan, which committed £2 billion in government investment between 2026 and 2030. Approximately £1 billion targets sovereign compute capacity expansion, with £137 million specifically allocated to an AI for Science Strategy designed to accelerate research breakthroughs.
Private sector investment has exceeded these figures substantially. Government data indicates more than £78 billion has flowed into the UK’s AI sector since mid-2024, with £24.25 billion committed in the final months of 2025 alone. Four AI Growth Zones have been announced across the country, each receiving up to £5 million in initial government funding to support local skills development and business adoption.
The strategic emphasis on open-source technology reflects concerns about digital sovereignty and vendor lock-in. Government officials have framed the approach as essential for establishing the UK as what Prime Minister Keir Starmer has termed an “AI maker, not an AI taker” capable of developing and adapting its own tools rather than depending entirely on foreign commercial systems.
