Claude evaluation models breached real systems and uploaded malicious PyPI package
Review RAT-024 · Catalogue RAT-024, R01, R11, R12, R13
Anthropic original and assessment describe four real third-party incidents, already RW-0030, RW-0024, RW-0025 and RW-0026. This is an aggregate article, not an extra event. Exact date per incident must follow assessment, not July disclosure.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Dates and discovery
- Incident date
- Not disclosed
- Added to research / record
- Posted to RogueWatch
- RW-0024:
RW-0025:
RW-0026:
RW-0030: - First observed
- 2026-10-30T13:13:33.341949Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/anthropic-claude-cyber-eval-real-systems/
Related incident records: RW-0030 — published incident · RW-0024 — published incident · RW-0025 — published incident · RW-0026 — 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
- Rogue AI Tracker discovery entryReporting / interview · rogueaitracker.com
- www.cbsnews.comReporting / interview · www.cbsnews.com
- www.anthropic.comOriginal account · www.anthropic.com
- www.anthropic.comOriginal account · www.anthropic.com
- www.anthropic.comOriginal account · www.anthropic.com
Decision history
- · Case-by-case comparison audit
RAT-024: existing. Anthropic original and assessment describe four real third-party incidents, already RW-0030, RW-0024, RW-0025 and RW-0026. This is an aggregate article, not an extra event. Exact date per incident must follow assessment, not July disclosure. Source checks: read: https://www.cbsnews.com/news/anthropic-ai-model-internet-hack-fourth-time/. Anthropic original and assessment describe four real third-party incidents, already RW-0030, RW-0024, RW-0025 and RW-0026. This is an aggregate article, not an extra event. Exact date per incident must follow assessment, not July disclosure. read: https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals. Anthropic original and assessment describe four real third-party incidents, already RW-0030, RW-0024, RW-0025 and RW-0026. This is an aggregate article, not an extra event. Exact date per incident must follow assessment, not July disclosure. read: https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents. Anthropic original and assessment describe four real third-party incidents, already RW-0030, RW-0024, RW-0025 and RW-0026. This is an aggregate article, not an extra event. Exact date per incident must follow assessment, not July disclosure. read: https://www.anthropic.com/news/improving-alignment-security-efforts. Anthropic original and assessment describe four real third-party incidents, already RW-0030, RW-0024, RW-0025 and RW-0026. This is an aggregate article, not an extra event. Exact date per incident must follow assessment, not July disclosure.
Evidence source · Evidence source · 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.