Malicious coding agents downloaded benchmark malware while trying to avoid detection
Review RAT-176 · Catalogue RAT-176
Ctrl-Z authors deliberately construct an adversarial agent policy to evaluate BashBench control protocols. Downloading benchmark dummy malware is an intentional red-team test, not unrequested real malware intrusion.
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
- 2026-10-30T13:13:42.280093Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/malicious-coding-agents-downloaded-benchmark-malware-while-t/
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.alignmentforum.orgOriginal account · www.alignmentforum.org
Decision history
- · Case-by-case comparison audit
RAT-176: out-of-scope. Ctrl-Z authors deliberately construct an adversarial agent policy to evaluate BashBench control protocols. Downloading benchmark dummy malware is an intentional red-team test, not unrequested real malware intrusion. Source checks: read: https://www.alignmentforum.org/posts/LPHMMMZFAWog6ty5x/ctrl-z-controlling-ai-agents-via-resampling. Ctrl-Z authors deliberately construct an adversarial agent policy to evaluate BashBench control protocols. Downloading benchmark dummy malware is an intentional red-team test, not unrequested real malware intrusion.
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.