These AI discussions, a notable outcome of the Trump-Xi summit held in May, are anticipated to occur prior to Chinese President Xi Jinping’s scheduled visit to the US on September 24, according to anonymous sources. However, the specific dates have yet to be confirmed.
China and the US, as economic, political, and military competitors, are increasingly wary of each other’s advancements in AI technology within the context of a rapidly escalating global AI race.
Both Beijing and Washington are exploring regulatory measures for advanced AI models that could enhance military capabilities, facilitate cyberattacks on essential infrastructure, and disrupt job markets.
The September discussions, initially reported by Reuters on Tuesday, will be spearheaded by Treasury Secretary Scott Bessent from the US side, as four sources indicated.
The identities of additional participants from both nations, along with the agenda and venue for the discussions, remain undecided as preparations are still in the preliminary phase, three sources shared.
The White House, US State Department, and US Treasury did not respond promptly to requests for comments from Reuters.
China’s National Development and Reform Commission, along with its cyberspace regulator, industry and foreign ministries, and the State Council, also did not provide immediate responses to inquiries.
Bessent, when asked about the upcoming talks during an interview on Fox Business Network, did not confirm the meeting directly but highlighted concerns regarding intellectual property, stating that models should adhere to uniform standards.
“We have identified watermarks from our US large language models in numerous Chinese models, which is unacceptable,” he remarked on the US television network. “We will be examining this in the upcoming days or weeks.”
These discussions will mark the first official AI dialogue under Trump
Former US President Joe Biden and Xi previously agreed that decisions regarding nuclear weapon usage should be controlled by humans, not AI, following working-level discussions in Geneva in 2024. However, the forthcoming meetings will be the inaugural official US-China dialogue on AI during President Donald Trump’s administration.
The US has already restricted exports of certain advanced AI chips to China but has refrained from placing Chinese AI startup DeepSeek, memory chipmaker CXMT, and over 100 other companies flagged as national security threats on a trade blacklist, aiming to prevent further escalating tensions with Beijing, as Reuters reported.
In response to inquiries about potential trade actions, US Trade Representative Jamieson Greer noted that multiple US government agencies are reviewing the matter. He mentioned concerns regarding forced technology transfers, intellectual property issues, and energy subsidies, but did not provide any further specifics.
“We are closely examining how China is advancing its AI development to ensure our companies can compete … on equitable terms,” Greer stated in an interview with CNBC.
Meanwhile, Chinese officials are considering restrictions on foreign access to the most advanced Chinese-developed AI models, amidst concerns regarding the hacking capabilities of Anthropic’s Mythos, which is currently not accessible to the public.
In May, Bessent indicated that future discussions would focus on “preventing the proliferation of powerful AI models … to non-state actors” and that meetings could commence “within the next four to eight weeks.”
After Trump’s visit, China’s foreign ministry confirmed that the two nations had agreed to initiate intergovernmental AI discussions, though Beijing has not since commented publicly.
China seeks engagement, rather than political discourse
China’s priorities for the upcoming talks include discussions on Mythos, controlling the release of future advanced frontier models by the US, and the potential restrictions Washington may impose on Chinese open-weight models, analysts indicated.
The Trump administration is consulting with US AI companies about creating voluntary safety standards for the launch of new AI models, although no regulatory measures have been announced so far.
Chinese authorities will continue their internal deliberations on AI topics over the coming month before determining which officials from various ministries will take part in the AI discussions, one source indicated.
China is seeking meaningful engagement and technical discussions on AI issues from the US, rather than mere political dialogues, the source elaborated.
“This inaugural meeting is unlikely to resolve any significant issues. Instead, they will aim to reach an agreement on basic definitions, such as what constitutes a frontier AI model,” suggested Paul Triolo, a partner at DGA-Albright Stonebridge Group.
