People’s Daily rejects US claims of malicious AI distillation, warns of countermeasures

The People’s Daily has dismissed US accusations that Chinese artificial intelligence firms engaged in large-scale replication of American cutting-edge models, warning of retaliatory measures if Washington persists in limiting China's AI industry. A commentary published on Wednesday by the official Communist Party paper labeled the claims as lacking factual or legal foundation and accused the US of attempting to politicize a standard technical and commercial activity.

The firm's uncompromising tone arrived just a week before the anticipated summit between President Xi Jinping and US President Donald Trump on September 24, which could feature AI as a central issue of discussion. Speaking in the commentary, the paper urged both sides to "strive for a shared objective and manage risks through dialogue and communication", while clarifying that Beijing would employ "countermeasures" if Washington exploited the allegations to impede Chinese enterprises.

In the realm of machine learning, distillation refers to a technique where a smaller "pupil" model learns from the outputs of a larger "instructor" model. Although this method is neutral and frequently employed, the commentary pointed out that some American entities had utilized distillation to construct their own systems, without explicitly naming any particular organization.

One instance cited is Thinking Machines Lab, co-founded by former OpenAI chief technology officer Mira Murati. The firm's debut model, Inkling, debuted in July and utilized other models - such as Moonshot AI's Kimi K2.5 - to generate initial training data. The People’s Daily article underscored that AI "ought not be the exclusive domain of major nations", and that the US was extending its restrictive policies from semiconductor controls and computing power limitations to AI model accessibility.

This move, the paper asserted, was driven by "deep-seated concerns regarding the potential erosion of its technological supremacy". The commentary responded to allegations raised last week by three US governmental bodies, which alleged that six Chinese corporations – DeepSeek, Moonshot AI, Alibaba Group Holding, MiniMax, StepFun, and Z.ai – had carried out "aggressive, malicious, and targeted" distillation campaigns of advanced American models, potentially with Chinese consent.

Alibaba, a major shareholder in the South China Morning Post, was among the implicated entities. The US Treasury Secretary, Scott Bessent, had previously stated that China could "never surpass" the US in AI due to the distillation process. AI executive Dario Amodei of Anthropic also connected the issue to the geopolitical rivalry between the two countries, advocating for a crackdown on unauthorized distillation by firms in authoritarian regimes and asserting that a Chinese leadership in AI would present a "serious danger" to the United States and the world.

Anthropic disclosed several Chinese businesses, including DeepSeek, Moonshot, Z.ai, and Alibaba, as companies it claimed had performed distillation of its models. In its latest report, Anthropic contended that these Chinese companies were participating in "unlawful" distillation due to the "industrial" volume of their attempts to access the outputs of Anthropic’s models - which the company prohibits from being used in China - via fraudulent accounts and fraudulent credit cards.

In a recent statement, the Chinese Foreign Ministry spokesperson, Guo Jiakun, refuted the notion that China's advancements in AI posed a threat to US national security. "The development of AI concerns the shared welfare of all humanity", Guo remarked during a routine press conference, urging nations to promote AI that is open, inclusive, broadly advantageous, and oriented towards the common good.

He warned that spreading alarmist narratives and pursuing confrontational or adversarial competition would only disrupt international AI governance and serve no one's interests.