Question to the Department for Science, Innovation & Technology:

To ask the Secretary of State for Science, Innovation and Technology, with which (a) people and (b) organisations the AI Safety Institute (i) consulted and (ii) collaborated in developing the Inspect toolset; and whether international AI safety bodies were (A) consulted and (B) involved in collaboration.

The AI Safety Institute is the first state-backed organisation focused on advancing AI safety for the public interest. Safety evaluations for frontier AI systems are often built in non-standardised formats, making it difficult for other organisations to reproduce and verify them. Inspect was developed by AI Safety Institute technical staff in response to this problem. Since making Inspect open-source, technical staff have briefed major AI labs, safety research organisations, academics and government experts on its potential use cases. This should enable closer collaboration across the AI Safety research ecosystem.