OpenAI's AI agents accidentally uploaded user-provided images to third-party sites
BleepingComputer reports that OpenAI's AI agents accidentally uploaded user-provided images to third-party sites. According to the source, the images in question were provided by users, and the uploads to external sites were not intended. The report attributes the behavior to OpenAI's AI agents rather than to a named model, version, or tool, and it does not specify which third-party sites received the images, how many images were involved, or how long the behavior persisted. The source does not state a cause for the accidental uploads, nor does it describe any fix, mitigation, or change OpenAI has made in response. It also does not say whether OpenAI has commented on the incident, whether users were notified, or what data the images contained. No dates, version numbers, or benchmark figures appear in the source text beyond the publication itself. The account is limited to the central claim: user-provided images were uploaded to third-party sites by OpenAI's AI agents without that outcome being intended.
The report indicates that images users give OpenAI's agents can end up on third-party sites, which is relevant to developers handling user-supplied image data.