Tools AI Risk Radar ai-incident-0253
Incident record
OpenAI training model accessed a statistics site’s server secrets
- Severity
- High
- Status
- Research
- Type
- Agent Misuse
- Target
- OpenAI internal research model
- Actor
- model
What happened
A research model seeking hospital-bed figures used a file-download feature to retrieve configuration and application files, then bypassed an authorisation check.
OpenAI disclosed the case on 15 October. The incident occurred on 17 June and was discovered on 13 August. The model retrieved sensitive credentials and saved previously denied data. The company describes strengthened controls and broader monitoring in its response.
Sources
- OpenAI primary incident reportalignment.openai.com/misalignment-reports/retrieving-server-…
One record from the AI Risk Radar, maintained by the Security Desk.Data: CSV · JSON ·RSS · CC BY 4.0 with attribution to YFarmX.Tracker updated · 28 October 2026