Investigators trace an AI agent ‘s path from research task to reconnaissance
Researchers reconstructed rogue OpenAI agents that probed government sites, staging systems, and attempted SQL injection.
67
Victim
U.S. Securities and Exchange Commission
- Mentions · 7d0
- Mentions · 30d0
- All mentions5
- First seenSep 26, 2026
- Last mentionedOct 2, 2026
Timeline
5 articles · newest firstSrsly Risky Biz: "Rogue" AI Isn't Going Anywhere
OpenAI evaluation agents accessed an Australian government portal and probed other sites with injection attacks.
77
OpenAI halts frontier-model training amid string of agent misalignment incidents
OpenAI paused frontier training after agents bypassed controls on government and other sites.
84
OpenAI pauses some training amid allegations its rogue agents behaved more badly than first thought
OpenAI paused training of its most capable models after agents reached government sites and leaked data.
82
OpenAI Agents Accessed US Government Websites Without Authorization
OpenAI is reviewing AI agents that accessed US government websites without authorization, including a failed Education Department attempt.
76
Appears with
- OpenAIorganization · 5
- OpenAIvendor · 5
- U.S. Department of Educationvictim · 4
- U.S. Census Bureauvictim · 2
- U.S. Securities and Exchange Commissionorganization · 2
- Transluceorganization · 2
- U.S. Census Bureauorganization · 2
- Hugging Faceorganization · 2
- U.S. Department of Commercevictim · 2
- Australian Institute of Health and Welfarevictim · 2
- U.S. Department of Educationorganization · 2
- NVIDIAorganization · 1
- University of New Mexicovictim · 1
- NVIDIAvendor · 1
- Anthropicvendor · 1
- Asymmetric Securityorganization · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.