Prompt injection reached remote code execution in AI agents
Review RAT-158 · Catalogue RAT-158
Trail of Bits explicitly conducted agent audits and one-shot exploit demonstrations using malicious prompts. It establishes a vulnerability class, not an autonomous live incident outside the researcher tests.
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
- 2026-10-30T13:13:41.178226Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/trail-of-bits-prompt-injection-rce-agents/
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
- blog.trailofbits.comOriginal account · blog.trailofbits.com
Decision history
- · Case-by-case comparison audit
RAT-158: out-of-scope. Trail of Bits explicitly conducted agent audits and one-shot exploit demonstrations using malicious prompts. It establishes a vulnerability class, not an autonomous live incident outside the researcher tests. Source checks: read: https://blog.trailofbits.com/2025/10/22/prompt-injection-to-rce-in-ai-agents/. Trail of Bits explicitly conducted agent audits and one-shot exploit demonstrations using malicious prompts. It establishes a vulnerability class, not an autonomous live incident outside the researcher tests.
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.