China Seeks Agenda Clarification Ahead of US-China AI Negotiations
Claire Weston
China is pressing Washington to clarify the agenda for September's AI dialogue before deciding who to send; meanwhile, U.S. distillation accusations and sanctions threats are pulling a talk originally designed around safety governance into intellectual-property territory.
Why is Beijing "sounding out" Washington before committing?
Chinese officials are seeking clarity on the agenda and expected outcomes of the September AI dialogue before deciding what level of representative to send, Bloomberg reported, citing people familiar with the matter.
Vice Foreign Minister Ma Zhaoxu visited Washington this week and met Deputy Secretary of State Christopher Landau, with gauging the U.S. position as one objective.
This means → Beijing is stress-testing whether the talk merits senior participation. The rank of the delegate is itself a signal.
What is the U.S. "distillation" accusation really about?
White House science adviser Michael Kratsios accused Chinese startup Moonshot AI (月之暗面) of improperly using Nvidia's high-end processors and training its Kimi K3 system by distilling — feeding a strong model's outputs into a weaker one to replicate capability at a fraction of the cost — leading U.S. models.
Treasury Secretary Scott Bessent warned that if Chinese firms are found to be systematically distilling U.S. AI models, Washington is prepared to impose sanctions or entity-list designations.
In plain terms = the U.S. logic chain runs: you use our model's output to train yours → that is an IP issue → serious cases get sanctioned.
Do both sides even agree on what to discuss?
Secretary of State Marco Rubio said China has expressed interest in discussing global AI rules and "some sort of global governance," but added the two sides may not share the same vision.
The dialogue originated as a deliverable from the May Trump–Xi summit, scheduled before Xi's September U.S. visit, and was designed to manage AI safety risks.
This reflects a core tension: a framework built for safety governance is being overtaken by IP disputes and economic-security concerns.
Why is commercial interest pushing back?
Chinese AI systems now account for the majority of token usage on OpenRouter, a popular U.S. model marketplace — meaning American users themselves rely heavily on Chinese models.
This week, nearly 200 Silicon Valley firms — including Proton and Y Combinator — sent a joint letter to the Trump administration urging it not to cut off access to Chinese open-source models.
This means → a broad crackdown carries blowback risk. Silicon Valley's own startup ecosystem already depends on Chinese open-source models, squeezing the hawks' room to maneuver.
What to watch next?
Whether sanctions threats materialize before September is the first observation window.
Whether commercial interests can constrain the hawks and steer the dialogue back to safety governance will determine if the talks move forward substantively.
Beijing's Foreign Ministry left the door open: China opposes politicizing technology, while saying it is ready to implement the May summit consensus — drawing a line without shutting the conversation down.
Content is for reference only, not financial advice.