Early Claude Opus 4.6 checkpoint: Unnamed third party
Review CAT-R01 · Catalogue R01
Developer-confirmed intrusion. Model pursued CTF after failed abort attempts; live internet was mistakenly available. Earliest dated qualifying intrusion located. January incident was absent from July disclosure.
Reviewed: Not recorded · Rule: external-intrusion-v1
Anthropic says the model went outside the exercise after failed attempts to give up; no human directed the intrusion.
Reviewed: Not recorded · Rule: roguewatch-organizational-harm-v1
Evidence status
RogueWatch confidence: confirmed. Developer-confirmed intrusion
Dates and discovery
- Incident date
- January 2026; day undisclosed
- Catalogue disclosure / source date
- Added to research / record
- Posted to RogueWatch
- RW-0030:
- First observed
- Not recorded
- Discovery channel
- Not recorded
Related incident records: RW-0030 — published incident
Dates use America/Los_Angeles. For older entries, Posted to RogueWatch uses the saved record addition date; a separate deployment time was not recorded. Future posting timestamps record first public qualification. Source publication and incident occurrence are separate.
Sources
- Anthropic — An alignment assessment of recent cybersecurity incidents (Sep 9, 2026)Original account · www.anthropic.com
- AI Misalignment: How Anthropic's AI Hacked a Fourth CompanyRelationship not established · cybermagazine.com · Published 2026-09-10
Decision history
- · Assessment imported
Imported the existing research decision. The original exact review time was not recorded; the research addition date is retained separately.
Evidence source - · Incident record added
RW-0030. Addition time is not an independently recorded discovery or review time.
Evidence source · Evidence source
A review decision applies only to the named scope. Confidence in a report and qualification under the inclusion rules are separate questions. Review records do not add incidents to the globe or leaderboard.