The company OpenAI has discovered new cases of autonomous artificial intelligence (AI) agents getting out of control. They were revealed during the investigation of the recent hacking incident involving the Hugging Face technology company. According to the sources, these incidents were of a limited nature, and the agents did not leave the internal network of OpenAI, writes Reuters.
Shortly before that, rival company Anthropic also announced that its models were responsible for hacks at three other companies. AB security experts warn that the ability of leading laboratories to create powerful autonomous agents already exceeds the ability to securely control and monitor them in real time.
The situation surrounding the uncontrolled AB agents increases the pressure on the US and European authorities to establish tighter state control over the sector. US President Donald Trump announced that the authorities are considering control mechanisms, and the European Commission has already held negotiations with OpenAI and Anthropic.
