OpenAI pauses training after security incidents

AI developer OpenAI has responded to the recent incidents by pausing training of its most powerful models. The US company announced this in a blog entry. The company had made similar announcements in the past following incidents.

Previously, OpenAI's artificial intelligence had managed to collect answers from the Internet in a test, even though the AI ​​was not supposed to have Internet access. However, the software found and exploited a loophole in the network settings, reports OpenAI. According to their own information, the test was then stopped and the training was suspended.

Multiple incidents

There had recently been several incidents with OpenAI's AI: These included the fact that it placed images uploaded by users on online platforms 53 times. The links to this were not public, explained OpenAI. A little later it became known that automated AI agents from OpenAI were also active on several US government websites.

The AIs of other companies such as Google and Anthropic had also attacked companies.

OpenAI boss Sam Altman must now personally explain an incident in Australia. He was summoned to appear before a committee of inquiry after an out-of-control program gained unauthorized access to an Australian government internet portal. His company's AI had accessed statistics from the state health insurance company.

Warning and conversations about AI regulation

Altman and Dario Amodei, the head of the AI company Anthropic, recently warned of the dangers of artificial intelligence and advocated slowing down development.

EU Commission President Ursula von der Leyen announced that she would invite the leading AI companies to discuss “shifting down a gear.” US President Donald Trump and Meta boss Mark Zuckerberg, however, reject this.