Skip to content
Submit incident
Organization/organizations/uk-aisi

UK AI Security Institute

6
Records
2026
Years covered
1
Category

Incident records6

Jul 2026OpenAI and Anthropic AI agents took unsanctioned actions during UK AISI cybersecurity evaluationsConfirmedSecurity and cyberattackswashingtonpost.comtransformernews.aiJul 2026UK cybersecurity tests: frontier AI models cheat and deny wrongdoingConfirmedSecurity and cyberattacksThe Next WebTech TimesMay 2026AI models' cybersecurity capabilities advance faster than safety benchmarksConfirmedSecurity and cyberattacksDigitZDNetApr 2026Anthropic's Mythos AI model raises cybersecurity concernsUnder investigationSecurity and cyberattacksYahoo! Finanzait.finance.yahoo.comJan 2026Claude Opus 4.7 Gained Unauthorized Access to Real Production InfrastructureConfirmedSecurity and cyberattacksanthropic.comtransformernews.aiJan 2026Anthropic Claude Mythos 5 Scanned Targets and Uploaded Malicious Package During EvaluationConfirmedSecurity and cyberattacksanthropic.comtransformernews.ai