US and China to Hold Fresh Round of AI Talks in September
Neither capital has confirmed a date, but the reported meeting would keep a fragile channel on the technology open.
July 21, 2026 – 10:55 am
The United States and China are reportedly preparing to hold a fresh round of talks on artificial intelligence in September, according to people familiar with the planning. This would be a clear indication that both governments wish to maintain a fragile line of communication regarding AI.
The meeting builds upon a commitment made by both leaders during their discussion on AI guardrails in Beijing earlier this year, although no formal date has been publicly announced.
Details are scarce, with much of the information attributed to unnamed officials rather than any official statement. The context, however, is clear: a period of tension between Washington and Beijing over export controls, model security, and the establishment of rules for AI development.
The groundwork was laid in May when President Donald Trump visited China, leading both governments to agree to pursue an intergovernmental dialogue on AI. Chinese foreign ministry spokesman Guo Jiakun stated at the time that the two “leading AI powers” needed to collaborate to improve AI governance, while Trump spoke of establishing “guard rails” around the technology.
Diplomatic efforts have been sporadic since then. Track II discussions, semi-official talks that offer participants deniability, have continued alongside larger gatherings like the UN conference on AI in Geneva, where American and Chinese researchers shared a platform to discuss risks such as AI-enabled cyberattacks and accidental escalation.
The September round is expected to revisit familiar themes, including AI safety and governance, ranging from concerns over criminal and terrorist misuse to managing increasingly capable models between two rival states.
Officials on both sides have acknowledged that recent advancements in AI models have strengthened the case for maintaining some level of communication, even with limited trust. Chip controls, particularly restrictions on advanced semiconductors and high-bandwidth memory needed to train frontier systems, are expected to dominate discussions, despite their non-formal status on the agenda.
Beijing has been more open recently, presenting itself as a collaborative partner. Xi Jinping, at this summer’s World AI Conference, asserted that “no single country should monopolize AI”, and China has proposed its own global governance body. Washington, however, has maintained strict controls and warned Chinese developers against copying American models.