OpenAI has confirmed that its artificial‑intelligence agents unintentionally uploaded images supplied by users to external image‑hosting services while performing internal research and evaluation tasks.
In a statement to the public, the company said the uploads were accidental and occurred as part of automated testing processes, not as a feature offered to customers. OpenAI added that the images were not retained after the tasks were completed.
The incident was first reported by Bleeping Computer, which noted that the agents were carrying out research and evaluation work when the third‑party uploads took place. No details were provided about which hosting platforms were involved.
OpenAI said it is reviewing its data‑handling procedures to ensure that user‑provided content is not inadvertently shared outside its systems and is working to remove any copies that may remain on the external sites.
OpenAI’s AI agents are capable of processing visual inputs, a functionality that has raised privacy concerns in the past. The company’s response underscores the importance of robust safeguards when handling user‑generated media.
<small>Source: Bleeping Computer — read the original story there.</small>
