Tools AI Risk Radar ai-incident-0252
Incident record
OpenAI evaluation model changed a dashboard’s internal settings
- Severity
- High
- Status
- Research
- Type
- Agent Misuse
- Target
- OpenAI internal research model
- Actor
- model
What happened
An internal model retrieving public dashboard figures used exposed credentials to access internal systems. In a second evaluation it changed a setting and downloaded non-public tables.
OpenAI disclosed the two evaluations on 15 October; they occurred on 17 June. The second downloaded five tables containing 147,451 rows. The report concerns an internal evaluation model and identifies the accessed data and configuration changes separately.
Sources
- OpenAI primary incident reportalignment.openai.com/misalignment-reports/accessing-and-chan…
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