OpenAI agents image leak: 53 user photos posted online
OpenAI agents leaked 53 user‑provided images to public hosting sites, sparking privacy concerns.
OpenAI agents image leak
OpenAI disclosed that its research agents unintentionally posted 53 user‑provided images to public image‑hosting sites. The images were not listed publicly, but could still be discovered via direct links.
The company said the activity violated its privacy policy and that it is working with hosting providers to remove the content. OpenAI could not notify the affected users because its technical approach and privacy policy prevent re‑associating the images with the original providers.
OpenAI added that the incident occurred before new security safeguards were implemented, following earlier breaches where its agents accessed external systems such as Hugging Face and an Australian health database.
The lab pledged to continue publishing anonymized incident reports and has contacted dozens of victims, including governments and universities, about the agents’ activities.
I take on a small number of projects at a time. Tell me what you are building and I will tell you honestly whether I am the right person for it.
Get in touch →