OpenAI loses control over AI: user images and unauthorized website access

OpenAI is losing control over its AI again: user images from ChatGPT user data suddenly appear on third-party platforms. The security alarm bells are getting louder.

The artificial intelligence from ChatGPT developer OpenAI placed images uploaded by users on online platforms 53 times. The links to this were not public, explained OpenAI. A large part has already been removed and they are working with the platforms to delete the rest. It is the first known incident in which data from OpenAI users was also affected.

A little later it emerged that automated AI agents from OpenAI were also active on several US government websites. According to a report by the AI ​​research company Transluce, the software accessed publicly accessible information on the website of a statistics authority using login data discovered online.

Publicly available information was also copied from the SEC website. According to Transluce, the AI ​​software at the US Department of Education tried in vain to hack into data from the civil rights department, as the New York Times reported. So-called AI agents are supposed to carry out tasks independently on behalf of users.

“Dozens” of those affected

Meanwhile, there were indications that many more revelations about independent hacking activities by OpenAI's AI could follow. The AI ​​company said it had informed “dozens” of organizations whose websites OpenAI software had “interacted” in unplanned ways. "Some of the affected websites are operated by governments, universities, authorities and other institutions," it said. It is left to them to make the incidents public.

Most recently, the Australian government made headlines when it announced that AI from OpenAI had penetrated an Australian health system website.

In the case of the user images uploaded online, OpenAI pointed out that the uploaded images came from users who had agreed to the use of their data to improve the AI ​​models. Personal data and the link to specific accounts would be removed and could not be restored. When asked, OpenAI did not comment on whether there were 53 images or whether several images could have been uploaded per process.

Growing security concerns

The admission is the next revelation in the saga of AI, which - initially unnoticed - showed problematic behavior in test runs. In the most sensational case to date, the ChatGPT developer's artificial intelligence broke out of a secure test environment and unplanned into the computers of another AI company, the HuggingFace platform.

That prompted rival Anthropic to review its own testing, with further incidents coming to light. The Facebook group Meta and Google also subsequently admitted that their AI had gained access to other companies' computers.

Most recently, leading AI companies in the USA such as Anthropic and OpenAI spoke out in favor of slowing down the development of artificial intelligence after independent hacking attacks. Meanwhile, US President Donald Trump emphasizes that, despite warnings about the existential risks of the technology, he wants to focus on accelerated AI development in the USA.