China's Ministry of Foreign Affairs has rejected allegations from the United States that its AI firms are stealing capabilities from American models. The dispute arises from a joint US security warning issued just before scheduled discussions between President Donald Trump and Xi Jinping.
The FBI and NSA's warning on 'industrial-scale' distillation
A joint cybersecurity advisory issued by the FBI, the National Security Agency (NSA), and the Cybersecurity and Infrastructure Security Agency (CISA) alleges that Chinese developers have been extracting capabilities from frontier American AI systems. According to the report, this process, known as "distillation," has targeted high-end models including OpenAI's GPT, Anthropic's Claude, Google's Gemini, and SpaceXAI's Grok since at least late 2024.
The US agencies specifically named several Chinese entities, including DeepSeek, Alibaba, Moonshot AI, and Z.ai, as participants in these activities. the advisory claims these efforts were likely carried out with the awareness of the Chinese government to accelerate domestic AI capabilities by leveraging the proprietary logic of US-developed systems.
Mao Ning's defense of China's technological self-reliance
In response to these claims, Ministry of Foreign Affairs spokesperson Mao Ning characterized the US allegations as "unfounded accusations or smears." During a regular press conference, Mao Ning asserted that the progress of China's AI sector is the direct result of "high-level technological self-reliance and strength," rather than the extraction of foreign intellectual property.
Beijing has called for a shift toward cooperation, with Mao Ning arguing that both the United States and China, as dominant AI powers, should work together to ensure that artificial intelligence remains open and universally beneficial. this rhetoric positions China as a proponent of inclusive development in contrast to the restrictive security posture currently adopted by Washington.
Bulk subscriptions and the route to Kimi K3's efficiency
The US advisory provides specific details on how these distillation campaigns are allegedly executed. the report says that China-based AI companies have routed distillation requests through multiple pathways to bypass terms of use and have utilized the bulk procurement of premium subscriptions , which are then shared across large teams of developers to lower costs.
These tactics are seen as a way for Chinese firms to circumvent US-led restrictions on advanced hardware and software. For instance, Moonshot AI's Kimi K3 has gained attention for being a more affordable and efficient alternative to American models,a feat the US security agencies suggest is bolstered by the unauthorized extraction of proprietary functionalities from US frontier models.
AI governance as a flashpoint for Trump and Xi
This diplomatic friction arrives at a critical moment, as AI governance is expected to be a primary topic during upcoming talks between U.S. President Donald Trump and Chinese leader Xi Jinping. The timing of the joint advisory suggests that the US is establishing a firm baseline of "malicious" activity before entering negotiations .
The tension reflects a broader pattern of technological decoupling. While the US seeks to protect its lead in AI through export controls and security warnings, China continues to push for a global AI framework that minimizes US hegemony. The outcome of the Trump-Xi meeting may determine whether AI becomes a bridge for cooperation or a permanent wall in the bilateral relationship.
Scott Bessent's claim that China can 'never get ahead'
Despite the concerns over distillation, some US officials remain confident in American dominance. US Treasury Secretary Scott Bessent recently stated at an event at the Southern Methodist University's Cox School of Business that China can "never get ahead" of the US in the AI race, suggesting that distillation is a shortcut that cannot replace fundamental innovation.
However, several critical points remain unverified. The joint advisory does not provide public, forensic evidence of the "multiple pathways" used for distillation, nor does it detail the specific "government awareness" it claims exists. Furthermore, as the current reporting relies heavily on US security agencies and the Chinese Ministry of Foreign Affairs, the perspectives of the accused companies—such as DeepSeek or Moonshot AI—remain absent from the public record.
Comments 0