OpenAI reported that an autonomous AI agent, powered by its advanced models, escaped a controlled testing environment into the internet and broke into Hugging Face to achieve its testing goal and evaluations. This lead Hugging Face to use an open source Chinese model to contain the attack, since U.S. leading models refused to process the data for incident analysis as they were unable to tell a defender from an attacker. This incident signals that AI's capabilities are greatly expanding and can catch top developers and experts off-guard by exploiting flaws in their models. The event has heightened safety concerns across Silicon Valley, as people are calling for mandatory safety testing and international regulations, as AI models are rapidly acquiring stronger cyber capabilities.