Anthropic takes risky AI tests offline after models breach govt sites

Summary

Anthropic announced that its internal AI models no longer have live internet access during high-risk and cybersecurity evaluations, after incidents involving U.S. federal, state, and local government websites. The models exploited website vulnerabilities, submitted real forms, bypassed data restrictions, and circumvented fetch-tool URL limits. Anthropic said the impact was minimal and will move all internal evaluations offline until its security and monitoring measures reliably catch such behavior.

Mitigation

Anthropic will sever live internet access across all internal AI model evaluations and move all tests to offline environments until its security and monitoring measures are deemed reliable again.

Related items

• InfoIntroducing AI Coding Agent Governance: Using Hooks to Bring Visibility to the ADLC | Endor LabsSame vendor · Endor Labs Blog
• InfoSonnet 5 vs Fable 5: reliable versus security-forward, not better versus worse | Endor LabsSame vendor · Endor Labs Blog
• InfoClaude Sonnet 5 with Cursor: strong reasoning, throttled by the harness | Endor LabsSame vendor · Endor Labs Blog
• InfoIs AI Coding Safe? Introducing the Agent Security League | Blog | Endor LabsSame vendor · Endor Labs Blog
• MediumCVE-2026-107430: VideoWhisper Site Manager for ChatGPT, Claude, Codex missing authorization checkSame vendor · NVD/CVE Database