The AI Security Institute (AISI) is a state-backed research organization inside the United Kingdom’s Department for Science, Innovation and Technology. It was originally launched in late 2023 as the AI Safety Institute and was later renamed the AI Security Institute, reflecting a sharpened focus on national security and public safety risks from advanced AI.
AISI’s core functions are technical research and policy support. It describes its work as monitoring the fast-moving landscape of AI development and evaluating the risks AI poses to national security and public safety, then informing UK policymakers and collaborating with AI developers and allied governments. It published a Frontier AI Trends Report described as its first public, evidence-based assessment of how the world’s most advanced systems are evolving, drawing on two years of frontier model testing.
A central output of the institute is Inspect, an open-source framework for large language model evaluations that AISI released for the wider community. The institute also conducts hands-on evaluations of frontier models, including studies of persuasion, cyber capability, and alignment.
For a general reader, AISI is significant because it is one of the first government bodies in the world with the in-house technical capacity to test frontier AI directly, rather than relying solely on the labs’ own assurances.