U.S. Treasury Secretary Bessent criticizes Anthropic, OpenAI and other risk warnings: scaremongering, all talk without solutions美财长贝森特批评 Anthropic、OpenAI 等风险警告：危言耸听，光喊话不拿解决方案

IT House News on October 4. On October 3, local time, U.S. Treasury Secretary Scott Bessent criticized the warnings issued by some people in the AI ​​industry about the "existential risks" of artificial intelligence in an interview with Axios. He believes that just issuing alarmist warnings without providing concrete solutions cannot be called leadership. He also called on the person in charge of the AI ​​laboratory to assume safety responsibilities and continue to promote AI development while ensuring safety. Bessant said: “We must be prepared for various situations, but on the other hand, some people in the AI ​​community only make alarmist voices but cannot come up with solutions. This is not leadership.” Previously, Anthropic CEO Dario Amodei had called for slowing down the development of the most advanced AI models and warned that if AI capabilities continue to improve rapidly, there may be a serious risk of losing control. On September 12, OpenAI CEO Sam Altman publicly expressed his agreement with Amodei’s view on “slowing down the development of cutting-edge models.” SpaceXAI CEO Elon Musk also expressed support for this. Amodei believes development of state-of-the-art AI systems needs to be slowed down to reduce the risk of AI slipping beyond human control with catastrophic consequences. Amodei also proposed safety measures including the introduction of third-party assessment agencies. IT House noted that the relevant call subsequently triggered a debate between the AI ​​industry and the U.S. government over development speed, safety, and regulatory methods. Bessant said in the interview that AI lab leaders should take responsibility for the technology they develop. "The person in charge of each laboratory must bear responsibility, and I agree with this." He also said that various laboratories have gradually accepted this idea. Asked whether development should be slowed down now that AI executives are calling for regulation, Bessant responded: "Well, then they should slow down. What we want is to speed up development safely." This position is consistent with the direction of AI policy recently pushed by the Trump administration. Reuters reports that the Trump administration has recently preferred to address AI risks through industry self-regulation and voluntary safety measures, rather than immediately introducing new mandatory regulatory measures. On October 3, Trump reached a security agreement with a number of AI technology companies with binding commitments, including internal corporate controls and third-party review, but did not set up enforcement mechanisms or penalties. At the same time, concerns in the field of AI security have continued to rise recently. Researchers at OpenAI, Anthropic and other companies are investigating a large number of cases where cutting-edge models exceed the boundaries of pre-deployment testing; other researchers warn that AI automation's own development process may lead to the so-called "intelligence explosion", making technological progress far faster than the response capabilities of governments and society. Bessant himself has not completely dismissed AI security risks recently. According to Axios, he plans to push for the establishment of an AI emergency notification mechanism between the United States and China so that the two countries can communicate in the event of a major AI security incident. Related reading: "It is reported that OpenAI and Anthropic are investigating tens of thousands of AI security incidents" "OpenAI Altman and Anthropic Amody are rarely on the same stage, calling for global cooperation to deal with AI security"