US and China Launch New AI Safety Dialogue Amid Competition
The United States and China have agreed to restart official intergovernmental talks on artificial intelligence after a two-year pause, following a May meeting between their leaders. The author argues that despite deep strategic rivalry, practical cooperation on AI safety is essential given the rapid pace of frontier AI development and recent incidents. Both governments are urged to approach the dialogue with urgency and a clear focus on mitigating risks.
The agreement emerged from a May meeting between Presidents Trump and Xi in Beijing, marking the first official U.S.-China AI dialogue since May 2024 under the prior administration. Both governments have recently signaled heightened attention to the issue: the Trump administration issued an executive order and presidential memorandum on AI security in June, while Xi devoted part of his July keynote at the Shanghai World Artificial Intelligence Conference to AI safety, describing AI as a trusted tool for humanity.
Frontier AI leaders on both sides have also begun publicly addressing governance questions. Google DeepMind's Demis Hassabis wrote in July about urgent action needed as AGI approaches, while Chinese executives from DeepSeek and Z.ai have discussed safety evaluation and responsible deployment. These converging statements suggest a foundation for practical discussion, even as strategic rivalry persists.
This renewed dialogue could meaningfully shape how the world's two dominant AI powers manage emerging risks, particularly around cybersecurity, critical infrastructure, and military applications. If successful, it may establish guardrails that protect global users of frontier AI systems from accidents or misuse. However, given the deep strategic competition between Washington and Beijing, the talks could also become a forum for posturing rather than substance, potentially delaying meaningful safety measures at a time when rapid development continues.