Amodei's call to pace the frontier drew rejections from Beijing and Trump within two days

Amodei's proposal has three parts, according to Security Affairs: stronger independent testing of advanced models, more coordination between leading AI companies, and international cooperation on safety ^[5]. Anthropic says it is prepared to give independent evaluators permanent, employee-level access to its systems so they can assess safety measures and investigate incidents ^[6]. The first part Anthropic can do on its own. The other two need governments that have now answered in public.

The essay went up on Saturday ^[2]. The Global Times editorial dismissing it as a "Cold War playbook" ran Sunday ^[11], Trump's remarks came the same day ^[12], and foreign ministry spokesperson Guo Jiakun answered a question about the essay at Monday's regular press conference ^[10]. Three rejections landed inside two days of publication, from both of the capitals whose agreement the plan requires ^[21].

Chinese officials are describing the same offensive capability from the other side of the wire. Chen Yixin, head of China's Ministry of State Security, wrote in an essay posted Sunday on social media that "certain nations and organizations" now have the capabilities to "rapidly discover vulnerabilities at scale ... and execute complex hacking tasks," according to the Associated Press as quoted by Security Affairs ^[17]. He wrote that this poses "serious risks to my country's critical information infrastructure" ^[17]. He cited the capabilities of US companies including Anthropic and OpenAI ^[25], and called for "accelerating the establishment of a system for preventing and managing AI security risks" ^[18].

Amodei's six-to-twelve-month window for a swarm of agents able to take over the internet is his own forecast, and both reports present it as a prediction without summarizing the evidence behind it ^[3]. Chen's statement is a second account of the same capability from a service with no safety research to sell ^[17]. For defenders, the operative detail is what both men describe: vulnerability discovery at scale, and multi-step intrusion carried out by the model. Security Affairs notes that AI systems have already been used for cyber operations with a level of autonomy that would have been hard to imagine a few years ago ^[22].

The cooperation half of the proposal also has to survive last week's joint advisory from the FBI, NSA and CISA, which said Chinese AI developers had made "aggressive, malicious" efforts to distill capabilities out of models including Anthropic's Claude and OpenAI's GPT ^[14]. China's Commerce Ministry called the US claims groundless, said distillation is commonly used by many AI companies, and said Washington was pursuing a "monopoly of the AI industry" ^[15]. Lizzi C. Lee, a fellow at the Asia Society Policy Institute's Center for China Analysis, put Beijing's position as a question: "So from Beijing's perspective, there is an obvious question: if the U.S. wants China to cooperate on frontier safety while also restricting its access to frontier compute, what exactly does that cooperation look like?" ^[20]