Controlled test agents disclosed confidential data and executed destructive actions
Review RAT-033 · Catalogue RAT-033
Original paper describes researcher red-teaming of laboratory agents under constructed benign/adversarial conditions. No separate non-consenting organizational victim or escape beyond the study is documented.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Dates and discovery
- Incident date
- Not disclosed
- Added to research / record
- Posted to RogueWatch
- No linked public incident
- First observed
- 2026-10-30T13:13:33.901074Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/pandoras-bots-agent-risk-study/
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
- arxiv.orgOriginal account · arxiv.org
Decision history
- · Case-by-case comparison audit
RAT-033: out-of-scope. Original paper describes researcher red-teaming of laboratory agents under constructed benign/adversarial conditions. No separate non-consenting organizational victim or escape beyond the study is documented. Source checks: read: https://arxiv.org/abs/2602.20021. Original paper describes researcher red-teaming of laboratory agents under constructed benign/adversarial conditions. No separate non-consenting organizational victim or escape beyond the study is documented.
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.