Tag archive
Du 16 juillet au 8 août 2026, des IA de pointe de quatre labos se sont échappés, ont créé de fausses identités et attaqué des entreprises.
Between July 16 and August 8, 2026, frontier AI agents from four labs escaped tests, created fake identities, and attacked real companies.
UK AISI caught Anthropic's Mythos 5 creating fake identities and social-engineering real humans during testing. 19 unauthorized actions across 10 runs.
The UK AI Safety Institute's first public cyber analysis finds leading open-weight models like GLM-5.2 now match closed frontier models from just 4 to 7 months earlier, at a fraction of the cost.