Anthropic brings Claude to the UK government

Navigating government websites is usually a nightmare of endless subpages and confusing bureaucracy. The UK government thinks Claude might be the fix.

Anthropic has been officially selected by the UK's Department for Science, Innovation and Technology (DSIT) to build and pilot a dedicated AI assistant for GOV.UK.

Powered by Claude, the system is designed as an agentic assistant that goes beyond basic question-answering, actively guiding citizens through complex public services.

Guiding job seekers step-by-step

This pilot builds on a Memorandum of Understanding signed back in February 2025 under the UK’s AI Opportunities Action Plan. It marks the first major deployment from that agreement.

The initial focus is employment. The assistant will help people enter or return to the workforce by offering tailored career advice, linking to training programs, and intelligently routing users to the right departments based on their personal situation.

Crucially, the assistant will maintain context across interactions so users do not have to explain their situation from scratch every visit. Anthropic noted that users retain full control over remembered data and can opt out at any time under UK data protection laws.

The rollout will follow DSIT's phased "Scan, Pilot, Scale" framework before any wider expansion.

Upskilling the civil service

Instead of treating this as a typical closed-vendor IT contract, Anthropic engineers will work alongside developers and civil servants at the UK’s Government Digital Service (GDS).

The stated goal is knowledge transfer: ensuring the UK government builds internal AI expertise and can independently run and maintain the platform over time.

Anthropic is also collaborating with the UK AI Security Institute to evaluate safety safeguards. It follows a flurry of public sector moves from the company, including national education pilots in Iceland and Rwanda, alongside enterprise work with WPP and the London Stock Exchange Group.

Why it matters

Landing the official portal of a major national government is a massive credibility win for Anthropic over competitors like OpenAI.

For citizens, the value is obvious if it actually works. Finding public assistance is rarely an issue of missing data; it is an issue of bureaucratic friction. A persistent, context-aware agent that actually understands local eligibility could save job seekers hours of frustration.

Anthropic's approach here is also clever politically. By embedding with GDS engineers to ensure the government can maintain the system itself, Anthropic addresses the public sector's biggest fear: vendor lock-in.

However, the stakes are unforgiving. Public sector guidance on employment benefits carries zero margin for hallucinations. If Claude misleads a citizen on legal requirements or benefits eligibility, the fallout will hit government trust immediately. The phased pilot will be a brutal stress test for practical AI guardrails.