OpenAI admits its tools accidentally published ChatGPT users' images
OpenAI acknowledged on Friday that its artificial intelligence tools inadvertently published 53 images sent by users via the ChatGPT platform on the internet, in a data leak incident that also affected the company’s own user data, according to the French news agency AFP.
The San Francisco-based company revealed that its tools also accessed websites belonging to U.S. federal agencies, confirming information previously reported by The New York Times. However, it emphasized that the data collected was limited to publicly available information.
OpenAI clarified that the links to the 53 images were not publicly accessible and that the images were mistakenly published on web hosting sites. The company stated that most of the images have been removed in coordination with hosting service providers, while efforts continue to delete the remaining ones.
Breaking News App
Browse Tactical Equipment
Agricultural
The images had been sent from accounts whose owners had consented to the use of their data for developing the company’s models. Before being used, the data passed through a filter for personal data privacy. The company confirmed that the images can no longer be linked to the users who sent them.
OpenAI did not specify whether the images contained identifiable individuals or sensitive data, nor did it provide additional details in response to questions from AFP.
The company attributed the publication of the images to the behavior of “AI agents,” which are software built on artificial intelligence models capable of autonomously executing a series of tasks.
It explained that agentic AI tools used in its research transferred training data to external services, noting that most of that data did not originate from users and that the majority of these transfers occurred before security measures surrounding its research were strengthened in August.
The company confirmed that it is currently conducting a comprehensive review of the past activities of its agentic tools, a process expected to take several months.
Company CEO Sam Altman acknowledged on Friday that the company “was not as fast as it would have liked” in its review and incident disclosure processes, citing the massive volume of data that needs to be analyzed.
An OpenAI spokesperson told AFP that the majority of activities reviewed so far fell within the scope of routine research tasks, such as accessing publicly available content on the internet to answer questions.
The spokesperson added that some of these activities involved government websites, considered as reference sources for public information.
The company confirmed that its models accessed public information published on the websites of the U.S. Securities and Exchange Commission (SEC) before part of it was published on another internet page, emphasizing that it found no evidence of access to non-public information.
