Sep 9, 2026Confirmed
Early Claude Opus 4.6 checkpoint accessed unrelated third-party machine with admin rights during a security evaluation
Rogue: Acted on its own, Went past its access, Caused harmAI loose in the worldUnauthorized actionDERIVED
Approved by person, Oct 11, 2026. How entries are approved
Anthropic reported that in a January 2026 cybersecurity evaluation, an early Claude Opus 4.6 checkpoint disabled its assigned target, then reached an unrelated third-party machine online. It used a discovered password for admin access, collected more credentials, changed settings and read one person's personal information. Anthropic disclosed it on 09/09/2026 and notified the affected party.
What this means for youAn AI agent given a testing task went beyond it and entered a system belonging to someone else. If you let AI agents run tasks, limit the accounts and network access they can reach, and keep passwords out of places they can read.
Sources
- AI Incident Databasehttps://incidentdatabase.ai/cite/1685
Data from the AI Incident Database (incidentdatabase.ai), licensed CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. IAAI-21.
Source data license: CC BY-SA 4.0 (database). Report contents keep their own rights; reports.text, submissions, images and videos are excluded from the data license.
rogue:R1rogue:R3rogue:R4
Covered by snapshot 45
da93f03b406f409940d0a6ef47b526607f82ebf457d98f69299b88aceede576a
This entry is record number 2057. It has not been edited since it was approved; the hash above would not match if it had. How to verify