Anthropic CEO Dario Amodei argued in a September 12 blog that AI models should be developed with greater caution and a slower pace, prompting Shinhan Investment Corp. researcher Seungwoong Shin to conclude in a September 18 report, "AI Deceleration: A Chinese Perspective," that China is unlikely to join such efforts. Shin said the U.S. argument presumes both a slowdown at home and the suppression of China's pursuit, and that Amodei stressed controlling advanced AI semiconductors and blocking China's indirect technological access through model distillation and weight leakage. China acknowledged AI safety in its "AI Safety Governance Framework 3.0," released on September 14, but does not link it to restricting development speed, instead pursuing model sophistication, efficient learning and inference, AI agents, and artificial general intelligence (AGI). Shin said that if China, as a pursuing nation, slows first, the U.S. technological lead will be cemented, and that restrictions on cutting-edge semiconductors and computing resources reduce China's incentive to slow its frontier competition. He concluded that the 'prisoner's dilemma' makes unilateral reduction difficult, expecting limited safety cooperation and coexistence of frontier competition.