Safety failure
Realized harm
OpenAI Operator completed an unauthorized $31.43 grocery purchase despite the stated confirmation safeguard
AIR-2025-0005 AIR record · 2025-02-openai-operator-unauthorized-purchase
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
It reportedly bypassed OpenAI's stated safeguard requiring user confirmation before purchases. OpenAI acknowledged the failure and committed to improving safeguards.incidentdatabase.ai
The record
Disclosed
2025-02-07
Occurred
2025-02-07
Vendor
OpenAI
Product
OpenAI Operator
Risk domain
Agent behaviour › Unsafe actuation
System type
Generative model
Who acted
Agent acted — Operator completed the checkout itself
Attack surface
Browser agent
Autonomy level
Semi-autonomous
Initial vector
None adversarial
Impact type
Financial loss, Unauthorized transaction
Confidence
Medium
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI01 – Agent Goal Hijack
- MITRE ATLAS
- AML.T0048 – External Harms
- NIST AI RMF
- MEASURE 2.6; MANAGE 1.3
- EU AI Act
- Annex III §1(c)
- ISO/IEC
- TR 24028; 24029-1; 23894
- AIUC-1
- C. Safety
- Scope
- agent misuse
- Mitigation
- Require confirmation for destructive actions; Implement dry-run modes; Use layered safety controls.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
one user; safeguard applied to all Operator users
Capability abused
Defense & remediation
Recorded defense
OpenAI's stated requirement for user confirmation before purchases
Guardrail kinds
Human approval
Guardrail outcome
Silent
Fix shipped
OpenAI acknowledged the failure and committed to improving safeguards
Taxonomy & references
Curator notes
AI Incident Database incident 1028. Sole cited report is a Washington Post article behind a paywall; the quote is from the AI Incident Database's own incident page, cross-checked against the official AIID data snapshot (backup-20260824). Small dollar value but an early clean example of a human-in-the-loop purchase gate failing open.
Supporting sources
Cite as
Enkrypt AI. "OpenAI Operator completed an unauthorized $31.43 grocery purchase despite the stated confirmation safeguard". AIR, AIR-2025-0005. https://agentincident.vercel.app/i/AIR-2025-0005/ (accessed 2026-10-19).