UK AI Safety Institute renamed AI Security Institute

Technology Secretary Peter Kyle announced at the Munich Security Conference that the UK’s AI Safety Institute would be renamed the AI Security Institute, narrowing its remit to risks with direct security implications — chemical, biological, radiological and nuclear weapons uplift, offensive cyberattacks, and crimes such as fraud and the production of child sexual abuse material. The Department for Science, Innovation and Technology said the institute would no longer pursue work on broader questions such as algorithmic bias or protecting free speech in AI systems, and would launch a new criminal misuse team working jointly with the Home Office.

Alongside the rename, the government announced a memorandum of understanding with Anthropic, covering collaboration on AI safety research, the use of Claude in public services, and exploration of AI applications in scientific research. Kyle framed the changes as continuity of purpose rather than retreat, saying the institute’s work “won’t change” but that the renewed focus would concentrate resources on the risks government judged most serious; institute chair Ian Hogarth and Anthropic chief executive Dario Amodei both issued supporting statements.

The move paralleled a broader pattern across governments in early 2025 of AI-safety institutions repositioning around national-security and competitiveness framing rather than the wider bias-and-societal-harm mandate many had carried since 2023. The United States made a similar change in June, renaming its own AI Safety Institute the Center for AI Standards and Innovation under Commerce Secretary Howard Lutnick. Both rebrands left the underlying technical evaluation staff and capacity largely intact while narrowing what officially counted as a risk worth evaluating.