AI Safety Institute
A national-level body (UK AISI, US AISI, and analogues in Japan, Singapore, Canada, others) that evaluates frontier AI systems for safety-relevant capabilities pre- and post-deployment. Frontier model providers voluntarily submit to AISI testing under the Bletchley framework.
Related terms
- Bletchley Declaration
A November 2023 declaration signed at the UK AI Safety Summit by 28 countries plus the EU, committing to international cooperation on frontier-AI risk. Kicked off the network of national AI Safety Institutes now working with model providers on pre-deployment testing.
- Frontier Model
The current-generation top-tier foundation models — GPT-5, Claude Opus 4.7, Gemini 2.5 Pro, o3 in 2026. Distinguished by superior capability at the cost of latency and per-token price. Also the term used by AI safety regulators (UK AI Safety Institute, US AI Safety Institute) to scope which models fall under enhanced review.
- Red-Teaming
Structured adversarial testing of an AI system, either manually or via automated attack pipelines, to find prompt injection, jailbreak, harmful-output, and misuse vulnerabilities before release. Increasingly required by AI regulation for high-risk systems.
Related on this site
Where this fits
AI Safety Institute is part of the Evaluation & Governance vocabulary used in the Generative AI Maturity Framework. See the full glossary for the complete set of 149 defined terms, or take the free maturity assessment to see where your organisation stands.