The UK's Department for Science, Innovation, and Technology (DSIT) has selected Anthropic to build an AI assistant to modernize how citizens interact with government services.

The main goal is to overcome the common difficulties in integrating large language models (LLMs) into public-facing platforms, which are often stuck in the proof-of-concept phase. The project aims to operationalize a memorandum of understanding signed in February 2025 with Anthropic.

An Agentic System to Guide Citizens

The initiative focuses on implementing agentic AI systems designed to actively guide users through processes, rather than simply retrieving static information. This approach aims to bridge the gap between information availability and actual user action, a common problem in government portals.

Using an agentic system powered by Claude, the assistant will provide tailored support, maintaining context between different interactions. The pilot project will initially focus on the employment sector, helping users find work, access training, and understand available support mechanisms.

Data Sovereignty and User Trust

The implementation of generative AI will follow a phased approach, with iterative testing before wider rollout. Anthropic has ensured that users will retain full control over their data, in compliance with UK data protection laws. The UK AI Safety Institute will test and evaluate the models.

A crucial aspect of the partnership is the transfer of knowledge. Anthropic engineers will work alongside civil servants and software developers at the Government Digital Service to develop internal AI expertise. This approach aims to reduce reliance on external vendors, treating AI competence as a core operational asset.

For those evaluating on-premise deployments, there are trade-offs to consider. AI-RADAR offers analytical frameworks on /llm-onpremise to evaluate these aspects.