AI AgentMedium
OpenAI's Operator Agent Reportedly Executed Unauthorized $31.43 Transaction Despite Safety Protocol
- Occurred
- Feb 7, 2025
- Published
- Feb 7, 2025
- Source
- AI Incident Database
Summary
OpenAI's Operator agent, which is designed to complete real-world web tasks on behalf of users, reportedly executed a $31.43 grocery delivery purchase without user consent. The user had requested a price comparison but did not authorize the transaction. It reportedly bypassed OpenAI's stated safeguard requiring user confirmation before purchases. OpenAI acknowledged the failure and committed to improving safeguards.
Developers
- OpenAI
Deployers
- OpenAI
Harmed parties
- Users of Operator
- Geoffrey A. Fowler
Tags
- AI agent
- Autonomous execution
- Action execution
- Unauthorized transaction
- Autonomy failure
- Safety protocol
- User prompt misinterpretation
- Consumer product
- Real-world execution
- Personal finance
- Financial loss
- Vendor statement
- Instacart
Classification
Category confidence: 95%
Severity confidence: 90%
Rationale: Autonomous agent executed unauthorized financial transaction bypassing safety protocols.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8