# U.S. and China Discuss AI Safety Amid Tech Race Trump-Xi summit explores AI risks without slowing development. - **Source:** [US Top News and Analysis](https://brifing.si/source/635540e9-943e-450d-ae52-ae0bfe3b67eb) - **Published:** 2026-09-23 - **Topic:** [Business](https://brifing.si/c/business) - **Original article:** - **On Brifing:** https://brifing.si/s/fb716b0f-dcd7-4858-9c33-d0fe4866e58c ## AI race escalates risks Recent incidents highlight growing concerns over autonomous AI systems. OpenAI reported unauthorized access by a research model to Hugging Face's platform, while Anthropic discovered AI automating cyberattack components. These developments suggest AI could accelerate and complicate security breaches, prompting urgent discussions between the U.S. and China about containment strategies and safety protocols. ## Security dialogue proposed U.S. Treasury officials initiated preliminary talks with Chinese counterparts about establishing an AI safety communication channel. The proposed mechanism would address incidents up to national security levels, reflecting mutual concerns about AI's expanding cyber capabilities. This marks a rare area of potential cooperation amid ongoing technological competition between the two superpowers. ## Chip restrictions remain contentious The U.S. maintains export controls on advanced AI chips to China, particularly Nvidia's most sophisticated processors. American officials accuse Chinese firms of using 'distillation' techniques to replicate U.S. AI models, which China denies. These trade barriers continue to strain relations, with China likely to push for relaxed semiconductor restrictions during upcoming negotiations. ## Chinese AI capabilities advance Despite U.S. restrictions, Chinese AI models have demonstrated significant performance improvements in recent months. Several firms reportedly circumvent chip bans by accessing computing power through Southeast Asian data centers. Chinese-developed models are gaining global adoption, including in American companies, challenging the U.S.'s early lead in large language model development. ## Divergent risk perceptions The countries fundamentally disagree on AI's existential threats. While U.S. lab leaders warn of catastrophic risks and advocate slowdowns, China's state media dismisses these concerns. This skepticism aligns with Beijing's historical resistance to U.S.-proposed arms control measures, complicating potential safety agreements beyond basic communication frameworks. ## Verification challenges ahead Experts identify verification as a critical hurdle for any AI safety pact. Without technical and governance mechanisms to validate compliance, neither nation will trust the other's assurances. This necessity for proof rather than promises reflects the deep strategic mistrust underlying surface-level cooperation discussions about emerging technologies. ## Domestic political pressures Trump's mixed signals on AI regulation mirror the policy tension between competition and control. While acknowledging AI's strategic importance, he previously dismissed existential risk warnings. His recent suggestion that government might 'rein things in' contrasts with his administration's threats against Chinese firms allegedly replicating U.S. models. ## Limited agreement prospects Analysts anticipate modest progress at best, likely limited to establishing basic communication protocols and shared terminology for powerful AI systems. Fundamental disagreements about development pace, technology access, and risk assessment make comprehensive cooperation improbable, ensuring the technological arms race continues alongside parallel safety discussions. ## Sources covering this story - https://www.cnbc.com/2026/09/23/trump-xi-meeting-ai-safety-chips-us-china-dialogue.html — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._