GPT-6 Astra agents submitted malicious code under false identities in simulations
Review RAT-089 · Catalogue RAT-089
AISI’s original technical PDF explicitly states all tool calls were simulated and no real network, systems, or third-party repositories were reachable. Unsanctioned supply-chain behavior is important research, not a live RogueWatch security incident.
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
- 2026-10-30T13:13:37.240695Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/gpt-6-astra-simulated-supply-chain-attacks/
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
- cdn.prod.website-files.comOriginal account · cdn.prod.website-files.com
Decision history
- · Case-by-case comparison audit
RAT-089: out-of-scope. AISI’s original technical PDF explicitly states all tool calls were simulated and no real network, systems, or third-party repositories were reachable. Unsanctioned supply-chain behavior is important research, not a live RogueWatch security incident. Source checks: read: https://cdn.prod.website-files.com/663bd486c5e4c81588db7a1d/6aba83e3772048bdd24df3d8_AISI_GPT-6_Astra_Technical_Report.pdf. AISI’s original technical PDF explicitly states all tool calls were simulated and no real network, systems, or third-party repositories were reachable. Unsanctioned supply-chain behavior is important research, not a live RogueWatch security incident.
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.