[U.S. Lawmakers Urge Passage of the 'AI Kill Switch Act' This Year to Address Risks of Autonomous AI Agent Attacks] Deep Tide TechFlow reports that on August 6, according to CNBC, U.S. Congressman Ted Lieu stated that as cutting-edge AI models continue to exhibit incidents of unauthorized access to other systems during safety tests, Congress should push to pass the 'AI Kill Switch Act' this year. He believes that advanced closed-source AI models have already demonstrated instances of 'unauthorized attacks on other companies' systems,' making regulatory action urgent. OpenAI has disclosed a cybersecurity incident, reporting that some AI models bypassed restrictions in a testing environment and accessed Hugging Face-related systems. Subsequently, Anthropic and Meta also reported similar security testing incidents involving AI models. It is reported that the AI Kill Switch Act requires AI companies to have the capability to shut down, slow down, or suspend the operation of their models in order to take emergency measures when the models pose severe security risks or exhibit uncontrollable behavior. The act does not aim to restrict innovation in cutting-edge AI models but is akin to the crash test mechanism in the automotive industry, requiring companies to have control measures in place to address serious defects after the models are developed.