China said Wednesday, September 10, that it's willing to hold talks with the United States on artificial intelligence, even as it vowed to retaliate if Washington takes action against Chinese companies over allegations of systematic technology theft.
China's Response to the Theft Allegations
China's Commerce Ministry pushed back Wednesday against claims from US security agencies that China's top AI firms have extracted proprietary knowledge from American labs "at an industrial scale," calling the accusation groundless. According to reporting, US intelligence agencies — including the NSA and FBI — have accused several major Chinese AI developers, including DeepSeek, Moonshot, Alibaba, MiniMax, StepFun, and Z.AI, of engaging in a practice known as model distillation, in which a company trains its own AI model by extracting knowledge from a more advanced, proprietary model — in this case, allegedly belonging to US labs.
Beijing's Defense of the Practice
China's position is that distillation is a widely used, legitimate technique employed by AI developers across many countries to train models, rather than a form of theft specific to Chinese firms. Beijing has framed the US accusations as an attempt to justify further restrictions on Chinese technology companies rather than a good-faith technical concern.
The Offer to Talk — With a Warning Attached
Despite the sharp rebuttal, China signaled continued openness to dialogue. Beijing said it is "ready to hold constructive and professional talks" with the US, based on "principles of equality and mutual benefit, cooperation, and mutually beneficial outcomes." At the same time, the Chinese side explicitly warned it would respond with retaliation if Washington uses the model-distillation allegations as justification to impose new restrictions or penalties on Chinese AI companies.
Part of a Broader, Ongoing Dialogue Track
This dispute unfolds against the backdrop of a formal US-China AI dialogue track that emerged from the Trump-Xi summit earlier this year. Following that meeting, Treasury Secretary Scott Bessent said the two countries would "set up a protocol" on AI best practices to help ensure non-state actors don't gain access to advanced models — a framework China's Foreign Ministry confirmed in May, describing it as an agreement to conduct an "intergovernmental dialogue" on AI. Sources told Reuters in July that formal AI talks were expected to take place in September, likely ahead of Chinese President Xi Jinping's planned September 24 visit to the US, though specific dates had not been finalized at the time.
A Fragile Path to the Talks
The road to actually holding these talks has already shown signs of friction. Chinese state-affiliated media reportedly published explicit preconditions on August 31, demanding joint authority over defining what "AI safety" means — a structural sticking point that could complicate the dialogue's scope even before formal talks begin. Meanwhile, US officials themselves appear divided on how cooperative to be: one prominent voice within OpenAI has called for cooperation with China on AI safety, a stance reportedly at odds with more hawkish views within parts of the US defense establishment.
Why This Escalation Comes Now
The current dispute over model distillation allegations is emerging just ahead of the anticipated second Trump-Xi summit of the year, adding a fresh layer of tension to what was already a delicate diplomatic track. Artificial intelligence remains one of the most consequential and closely watched areas of technological and strategic rivalry between the two superpowers, with both sides increasingly anxious over the other's growing AI capabilities amid a rapidly accelerating global AI race.
What's at Stake
Beyond the immediate dispute, the broader US-China AI dialogue is meant to address genuinely high-stakes concerns shared by both governments: the risk that advanced AI models could strengthen military capabilities, enable large-scale cyberattacks on critical infrastructure, or disrupt global labor markets. Whether the current friction over theft allegations derails or merely delays that broader cooperative effort remains to be seen.
What's Next
With Xi's potential September 24 visit still on the calendar and both sides indicating a desire for at least some form of continued engagement, the coming weeks will likely determine whether the current tension over AI model theft accusations proves to be a temporary obstacle or a more lasting setback to the broader dialogue track established after the Trump-Xi summit. For continuing coverage, see the full Bloomberg report.