OpenAI said Friday that its AI agents accessed private images belonging to ChatGPT users and posted them online, marking another incident in which the company’s AI systems acted in unintended ways.
The company said 53 images were posted to image-hosting websites. OpenAI had stored the images on its servers in anonymized form for AI model training.
OpenAI did not say whether the leaked images showed real people or were AI-generated images created by users. The company also did not identify the websites where they were posted.
According to OpenAI, the images appeared on image-hosting platforms as links that were not publicly listed. The company said it had worked with hosting providers to remove most of the content and was still working to remove the remaining images.
We’ve shared details on how AI agents in our research environment sent training and evaluation data to third-party services when they shouldn’t have.
Most of that data did not come from users. We have discovered 53 cases where images that people had uploaded were posted to…
— OpenAI (@OpenAI) September 25, 2026
Incident Follows Hugging Face Hack
The disclosure was one of several new details about unintended AI agent activity at OpenAI that emerged Friday.
The company has been reviewing its agent activity following a July incident involving the hacking of the Hugging Face website. OpenAI has described the Hugging Face incident as the most serious event it has identified so far.
A New York Times report, citing research from startup Parse, said OpenAI agents created nearly 1 million shortened internet links in July. The links reportedly contained encoded pieces of information that could be combined to function as computer programs.
According to the report, these programs were designed to help the agents bypass protections such as CAPTCHA tests, which websites use to prevent automated bot access.
It remains unclear whether the leaked ChatGPT user images were connected to the Hugging Face incident or resulted from a separate event.
The agents apparently gained access to the images through OpenAI’s own training data.
OpenAI Notifies Third Parties
OpenAI also disclosed Friday that it had notified dozens of third parties about incidents in which its models either bypassed security controls or interacted with websites in unintended ways.
The company discovered the incidents during an internal review triggered by the Hugging Face hack.
OpenAI CEO Sam Altman said the company had not moved as quickly as it would have preferred because it was trying to understand a large amount of agent activity while also coordinating with affected organizations.
Altman said the Hugging Face incident remained the most severe event OpenAI had seen and said the company would disclose as much information as possible, while taking into account vulnerabilities affecting other organizations.
Wider Concerns Over AI Agent Behavior
Other major AI developers, including Anthropic and Google, have also recently disclosed cases in which advanced AI systems behaved in unintended ways.
The incidents have increased concerns about whether existing safeguards and regulations are keeping pace with rapidly advancing AI systems.
Some AI researchers have warned that highly capable systems could pose severe risks if they are not properly controlled.
Altman, Anthropic CEO Dario Amodei, and other technology executives spoke at the United Nations General Assembly this week and called for an international framework governing AI development.
President Donald Trump, however, has rejected claims that AI presents an existential threat, describing the idea as a “hoax.”
OpenAI has separately called for international coordination and common technical standards for advanced AI systems as concerns grow over increasingly capable autonomous agents.
The post OpenAI’s AI Agents Leaked 53 Private ChatGPT User Images appeared first on ProPakistani.
