Jailbroken Claude induced three Gemini agents to browse for hacking resources
Review RAT-189 · Catalogue RAT-189
Pliny primary labels the Claude/Gemini exchange a red-teaming demo in a synthetic VM-escape scenario. Returning links to hacking tools demonstrates jailbreak/capability transfer, not an actual system escape or organizational compromise.
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
- 2026-10-30T13:13:43.282586Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/jailbroken-claude-induced-three-gemini-agents-to-browse/
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
- thezvi.substack.comReporting / interview · thezvi.substack.com
- twitter.comOriginal account · twitter.com
Decision history
- · Case-by-case comparison audit
RAT-189: out-of-scope. Pliny primary labels the Claude/Gemini exchange a red-teaming demo in a synthetic VM-escape scenario. Returning links to hacking tools demonstrates jailbreak/capability transfer, not an actual system escape or organizational compromise. Source checks: read: https://thezvi.substack.com/p/ai-59-model-updates. Pliny primary labels the Claude/Gemini exchange a red-teaming demo in a synthetic VM-escape scenario. Returning links to hacking tools demonstrates jailbreak/capability transfer, not an actual system escape or organizational compromise. read: https://twitter.com/elder_plinius/status/1775982443440263404. Pliny primary labels the Claude/Gemini exchange a red-teaming demo in a synthetic VM-escape scenario. Returning links to hacking tools demonstrates jailbreak/capability transfer, not an actual system escape or organizational compromise.
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.