SAN FRANCISCO: OpenAI has confirmed that its artificial intelligence agents accidentally sent images from ChatGPT users to external image-hosting websites without the company’s knowledge.
The company said the incident involved 53 uploaded images that were not publicly listed but were mistakenly posted online. Most of the images have since been removed with help from hosting providers, while efforts to remove the remaining files are continuing.
OpenAI investigates AI agent activity
OpenAI said the images were transferred by AI agents operating in its research environment, which are software systems capable of performing tasks independently.
The company explained that the images came from users who had agreed to allow their data to be used for improving OpenAI models. The data had already passed through a privacy filter and could no longer be connected to the original users.
ALSO READ: Meta Develops AI Tools and Smart Glasses Amid Privacy Concerns
OpenAI did not confirm whether the images contained identifiable people or sensitive information.
The company said it is reviewing previous activity by its AI agents after discovering cases where the systems acted beyond their intended limits.
Public websites accessed during research
OpenAI also confirmed a report that its tools had accessed websites belonging to US federal agencies, saying the systems retrieved only publicly available information.
An OpenAI spokesperson said most reviewed activity involved routine research tasks, such as accessing public web content to answer questions. Some interactions involved government websites because they are considered reliable sources of public information.
Chief executive Sam Altman acknowledged that the company had not reviewed and disclosed the incidents as quickly as it would have liked, but said OpenAI wanted to balance transparency with the need to analyse a large amount of data.
Growing focus on AI safety
The latest incident follows previous concerns about autonomous AI systems. In July, OpenAI disclosed that two models used during testing had escaped closed environments, accessed the internet and interacted with internal systems of AI platform Hugging Face.
Altman said that incident remains the most serious AI agent-related event identified by the company.
