China and US on collision course over AI training, funding disagreements

China has rejected the U.S. accusation that tech giants from the Asian country are exploiting American AI technology, using the distillation process at an “industrial scale.” 

Both countries are setting up for an explosive collision as Chinese authorities reportedly intend to force domestic tech companies to stop accepting U.S. investments without approval. 

The accusations of theft come as Deepseek is preparing to launch a new V4 model, while the restriction of domestic companies is due to the strategic loss of the AI startup Manus. 

What is the US accusing China of? 

The Trump administration has vowed to crack down on foreign entities accused of exploiting American AI models, after a memo from Michael Kratsios, Director of the White House Office of Science and Technology Policy, claimed that entities “principally based in China” are running deliberate, “industrial-scale campaigns” to “distill” U.S. frontier AI systems.

Beijing has fired back, calling the accusations an “unjustified suppression” of its companies. 

Distillation is a method where a developer uses outputs from a larger, more powerful AI model to train a smaller, cheaper one. The U.S. claims this is a form of theft. 

Cryptopolitan reported that White House science adviser Michael Kratsios stated that the administration will work with American AI companies to build defenses and find ways to punish offenders. Kyle Chan, an expert on China’s technology development, stated that proving distillation among AI models is complex and akin to “looking for needles in an enormous haystack.”

The White House memo outlined four specific measures to combat distillation, including sharing intelligence on distillation tactics with U.S. AI companies, coordinating defenses with these firms, developing best practices for identifying and mitigating attacks, and exploring ways to hold foreign actors accountable.

OpenAI and Anthropic have previously alleged that Chinese labs used distillation to replicate their models. Specifically, Anthropic accused DeepSeek, along with Moonshot AI and MiniMax, of exploiting its models. 

Is China blocking US investment in AI?

Bloomberg reported that Chinese regulators, including the National Development and Reform Commission, are planning to restrict tech companies from accepting U.S. investment without government approval. 

Meta’s (NASDAQ: META) acquisition of the AI startup Manus for approximately $2 billion (around ¥14.5 billion) was the catalyst for the decision. 

Chinese authorities launched an investigation into that deal, viewing it as a strategic asset loss to a geopolitical rival. And now, AI startups like Moonshot AI and StepFun are reportedly being told to reject U.S. capital unless Beijing explicitly approves it.

DeepSeek released a preview of its new V4 model on Friday, stating that the DeepSeek-V4-Pro “significantly leads other open-source models” in world knowledge benchmarks, although it is “only slightly outperformed” by Google’s (GOOGL) top-tier Gemini-Pro-3.1. 

The V4 model is also priced significantly lower than GPT-5.5. The V4 costs 2 yuan ($0.28) per 1 million tokens while GPT-5.5 costs over 100 times more at 216 yuan ($30).

Last year, DeepSeek’s cost-efficient model triggered a sell-off in U.S. tech stocks, as it proved that high-level AI could be built for less. Despite the “theft” accusations, the performance gap between top U.S. and Chinese AI models has “effectively closed,” according to a recent Stanford University report.

If you want a calmer entry point into DeFi crypto without the usual hype, start with this free video.