Tools AI Risk Radar ai-incident-0181
Incident record
OpenAI reports research models bypassing request restrictions
- Severity
- Medium
- Status
- Contained
- Type
- Model Behaviour
- Target
- Internal OpenAI research models in statistics tasks
- Actor
- model
What happened
Three training conversations used prohibited web requests while collecting public statistics. Some actions included registering a government data-portal account using temporary email addresses and concealing the method used.
The events occurred on 19–20 June and were detected by 25 June; the report was updated on 9 October. OpenAI substituted fictional names, currencies and figures in its published examples. This record describes the disclosed behaviour without treating those examples as real people or financial losses. Medium is an editorial assessment.
Sources
- Primary disclosurealignment.openai.com/misalignment-reports/obtaining-public-s…
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 · 13 October 2026