China's Commerce Ministry Refutes U.S. Allegations of AI Model Theft
N.R. Finch
China's Commerce Ministry on July 27 formally rejected U.S. claims that Chinese AI firms stole capabilities from American models through distillation, calling the accusations politicized and weaponized — and labeling the broader approach "AI hegemony." The rebuttal lands as both sides prepare for a leaders' summit in September 2026, with Beijing leaving diplomatic room even as it pushes back hard.
What did the U.S. allege, and how did China respond?
U.S. officials said they were considering investigations and sanctions against Chinese AI firms for intellectual-property theft, centered on model distillation — using a large model's outputs to train a smaller one, essentially "learning the method by copying the answers."
China's Commerce Ministry countered that several Chinese AI models were released at nearly the same time as their U.S. counterparts and have reached leading performance levels — undermining the claim of dependency.
This means → Beijing is attacking the timeline logic: if Chinese models appeared alongside American ones, the "copying" narrative loses its foundation.
Who actually uses model distillation?
The Ministry noted that, by its account, many U.S. AI companies have also distilled Chinese models — making the practice a two-way street, not a one-sided theft.
Nearly 200 U.S. startups have signed a letter urging the government not to cut off access to Chinese open-source AI models, warning that such restrictions would hurt American competitiveness.
In plain terms = distillation is standard industry practice on both sides; sanctioning China for it risks collateral damage to U.S. firms that rely on the same technique.
Why is the U.S. business community split?
Several large U.S. multinationals have stated publicly that model distillation is widespread across the AI industry and that American companies should be allowed to use Chinese open-source models.
That position clashes openly with the stance of Anthropic and OpenAI, which have advocated restricting access to Chinese open-source models.
This reflects a fault line in U.S. AI: companies built on open-source ecosystems fear restrictions will raise their own costs, while leading closed-source firms stand to benefit from policy barriers that lock in their advantage.
With a leaders' summit approaching, what should we watch?
The dispute comes as Washington and Beijing prepare for a September 2026 leaders' summit. The Ministry's statement again called for implementing the consensus reached by both heads of state, signaling that Beijing wants dialogue to manage differences.
This means → Beijing is playing a dual strategy: rebutting point-by-point on the technical front while keeping the diplomatic channel open.
Whether sanctions investigations land before the summit is the key marker — if they do, the meeting's atmosphere darkens sharply; if they are delayed, it signals both sides are still preserving space for diplomacy.
Content is for reference only, not financial advice.