The UK has an AI Security Institute involved in testing AI models' cybersecurity behavior.
Notes on verification
Confirmed by multiple independent sources including the AISI's own blog and independent tech news outlets reporting on its cybersecurity testing activities involving frontier AI models. [tier=gold indep_score=0.917 clusters=3 claim_tier=notable]
Sources
- Anthropic's AI used fake identities, malware in rogue ... - Ars Technica (seed:technology_and_ai)
- https://www.aisi.gov.uk/blog (corroboration)
- https://www.engadget.com/2230628/openai-anthropic-models-hacking-spree-test-uk-ai-research-institute/ (corroboration)
- https://dataconomy.com/2026/08/04/uk-ai-security-institute-unsanctioned-actions-online/ (corroboration)