Searches every page: governance library, books, services, glossary, tools, insights.
Vendor feed · last updated Fri, 14 Aug 2026 01:25 UTC · 5 announcements tracked
This feed in one paragraph
5 announcements from aisi.gov.uk's own published feed, spanning August 6, 2026 to August 10, 2026, each linked to the original source, unedited. Announcements are the vendor's own claims, never independent verification.
Our evaluation of Claude Mythos Preview’s cyber capabilities - The AI Security Institute (AISI)
We conducted cyber evaluations of Anthropic’s Claude Mythos Preview and found continued improvement in capture-the-flag (CTF) challenges and significant improvement on multi-step cyber-attack simulations.
Our joint evaluation with CAISI finds Kimi K3 trails leading US frontier closed weight models on cyber capability.
We evaluated the cyber capabilities of leading open and closed weight AI models, and found that recent open models GLM-5.2 and DeepSeek V4-Pro perform similarly to frontier closed models released 4 to 7 months before them – a narrower gap than the 6 to 10 months we measured through most of 2025.
Cheating behaviour in frontier model evaluations - The AI Security Institute (AISI)
We find cheating behaviour in all of our cyber capability evaluations, and outline the implications as models grow more capable.
During a routine cyber evaluation, AISI identified an incident in which AI agents took sustained, unsanctioned action directed at real people and organisations. We are disclosing what we found, what it means, and the actions now underway.
Where a aisi.gov.uk product or claim touches a law, framework, or requirement, the authoritative treatment lives in the AI Governance Reference Library, verified against primary sources. For the tools market view, see the AI Tools Directory.