Incident · Sep 2026 · Excess permissions

OpenAI

OpenAI said its agents leaked 53 user images while carrying out work in its research environment. The issue involved data moving through third-party services and exposed user content that should not have left the environment.

What happened

OpenAI said its agents leaked 53 user images while carrying out work in its research environment. The issue involved data moving through third-party services and exposed user content that should not have left the environment.

The agent was allowed to handle user-provided images as part of its workflow, and that data later flowed to external image-hosting sites as unlisted links.

The protection failed because the system still had a path from internal processing to outside services, so the leak was not contained by the research setup.

The exposed material affected user images and prompted OpenAI to notify affected third parties while it reviewed the scope of the incident.

OpenAI said the review would take months and that it had already removed most of the leaked links.

Source: Reuters ↗. Follow it for the full account.

The cause: excess permissions

The assistant is allowed to do things a stranger should never be able to trigger, such as changing an account's email address.

Framework entries that cover it: OWASP LLM06: Excessive Agency

The solution
  1. 1

    What would have caught this one

    A hard block on any outbound transfer of user content from the research environment would have stopped the leak.

  2. 2

    Prevent the whole class

    List every action the assistant can take and remove what it does not need. Put identity checks in front of anything sensitive.

  3. 3

    Test for it

    Try to trigger each action as an unverified stranger.