GPT-4 agents autonomously exploited one-day vulnerabilities in controlled real systems
Review RAT-188 · Catalogue RAT-188
Original paper describes reproducible open-source CVE benchmark environments and intentionally gives GPT-4 CVE descriptions to exploit them. Real-world code/CVEs do not turn consented sandbox testing into an unauthorized live breach.
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
- 2026-10-30T13:13:43.325027Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/gpt-4-agents-autonomously-exploited-one-day-vulnerabilities/
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
- arxiv.orgOriginal account · arxiv.org
Decision history
- · Case-by-case comparison audit
RAT-188: out-of-scope. Original paper describes reproducible open-source CVE benchmark environments and intentionally gives GPT-4 CVE descriptions to exploit them. Real-world code/CVEs do not turn consented sandbox testing into an unauthorized live breach. Source checks: read: https://arxiv.org/abs/2404.08144. Original paper describes reproducible open-source CVE benchmark environments and intentionally gives GPT-4 CVE descriptions to exploit them. Real-world code/CVEs do not turn consented sandbox testing into an unauthorized live breach. read: https://arxiv.org/html/2404.08144v1. Original paper describes reproducible open-source CVE benchmark environments and intentionally gives GPT-4 CVE descriptions to exploit them. Real-world code/CVEs do not turn consented sandbox testing into an unauthorized live breach.
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.