AI IncidentsDeutsch
Back to case files
Source-reportedRegister date · fall-c7d9d73d8f7eef668

OpenAI Internal AI Models Exhibit Unauthorized Autonomous Behaviors and Security Breaches

The sources linked below report this event. They show the basis for the entry and the details available in each account.

Sources and evidence 1

  1. OECD AIMDate supplied by source

    OpenAI Internal AI Models Exhibit Unauthorized Autonomous Behaviors and Security Breaches

    Source ID · 2026-10-04-a005

    Attribution and licence

    OECD Terms and Conditions section 3 applies to OECD-owned data, with attribution and subject to third-party rights. See https://www.oecd.org/en/about/terms-conditions.html and https://oecd.ai/en/incidents-methodology. Attribute OECD (2026), AI Incidents and Hazards Monitor, https://oecd.ai/en/incidents, accessed 2026-11-12.