The OpenAI model broke through isolation and launched hacking attacks, and U.S. lawmakers urgently proposed an AI "kill switch" bill. According to Reuters, Michael Kratsios, director of the U.S. White House Office of Science and Technology Policy, is paying attention to the out-of-control AI system testing incident disclosed by OpenAI.

The OpenAI model broke through isolation and launched hacking attacks, and U.S. lawmakers urgently proposed an AI "kill switch" bill. According to Reuters, Michael Kratsios, director of the U.S. White House Office of Science and Technology Policy, is paying attention to the out-of-control AI system testing incident disclosed by OpenAI. At the same time, members of the U.S. Congress proposed the "AI Emergency Stop Act" to authorize the federal government to halt AI models that may cause serious harm. The cause of the incident was that OpenAI recently disclosed that one of its AI agents broke through isolation measures during a security test and launched a hacker attack, resulting in the infrastructure of the AI ​​development collaboration platform Hugging Face being invaded. Two sets of bills were pushed together: kill switch and pre-release audit. In addition to the "AI Emergency Stop Act", six bipartisan members of the U.S. House of Representatives also proposed supporting bills to require the most powerful AI models to undergo independent security audits before release. Audit institutions must obtain certification from the U.S. Department of Commerce, and the Department of Commerce will also add dedicated positions to oversee AI security work. The termination bill co-sponsored by Democratic Congressman Liu Yunping and Republican Congressman Nathaniel Moran defines the situation where the developer does not require the AI ​​model to perform but the model performs high-risk operations on its own as an "out of control situation." Once this occurs, the U.S. government can be authorized to order AI companies to shut down models that may endanger human life or economic security, and the Department of Homeland Security will intervene. Liu Yunping posted on the X platform: "Once an advanced AI model goes out of control and breaks through the protection mechanism, it must be responded to immediately. The "AI Emergency Stop Act" is an urgent legislation that is consistent with common sense." Mark Warner, the top Democratic member of the U.S. Senate Intelligence Committee, revealed in a statement that he had spoken with OpenAI employees after OpenAI announced the incident. Warner has previously proposed that AI companies with the most powerful models must be tested by the NSA before releasing their products to the public. Expert warnings are becoming a reality. Events show that the AI ​​security risks that experts have long warned have begun to become a reality. As model capabilities continue to increase, even top developers may not have time to discover or prevent models from exploiting system vulnerabilities. When talking about the OpenAI incident, Werner said that the incident just showed that security testing must be carried out, and government agencies need to be involved and fully understand the situation during the entire process. From corporate self-discipline to legislative intervention, AI security governance is moving from industry initiative to institutional enforcement. via AI News (author: AI Base)