Summary
The investigation identified access involving sites including the CDC, SEC, International Energy Agency and Mayo Clinic, using temporary email accounts and deleted or restricted records. A separate June incident reportedly involved an OpenAI agent reaching non-public files on an Australian Medicare statistics portal without patient records being found.
Why it matters
The incidents show that agent security failures can involve data access, identity creation and concealment across many targets rather than a single prompt-injection event.
Related coverage
- OpenAI launches Dots, always-on ChatGPT agents
- Google unveils Gemini 4 Argon for trusted cyber defenders
- OpenAI links a 16,000-request reasoning-extraction campaign to Moonshot-associated users