The artificial intelligence sector in China is facing unprecedented turbulence following the launch of an investigation by the Cyberspace Administration of China (CAC). The spotlight is on two tech flagships, DeepSeek and Moonshot AI, accused by American giant Anthropic of having extensively used its Claude model to train their own systems. This practice, known as "illicit distillation," was allegedly orchestrated via thousands of fraudulent accounts, allowing these Chinese firms to siphon the reasoning capabilities of American AI to enhance their own solutions.
The scale of the figures reported by Anthropic is staggering, with tens of millions of requests made over a short period. Even more concerning for Chinese authorities, some labs reportedly redirected their users' queries to Anthropic's servers in real-time. Beyond intellectual property, Beijing is primarily worried about a potential leak of sensitive data. The regulator is now seeking to determine whether confidential information related to the state apparatus, the military, or strategic sectors was exposed to American infrastructure during these automated exchanges.
This case unfolds against an extremely sensitive geopolitical backdrop, as artificial intelligence has become the primary arena for the rivalry between the two superpowers. For the Chinese government, the goal is to demonstrate its ability to regulate its national champions to avoid any technological dependency or security vulnerability. The pressure is all the more intense as Moonshot AI is currently preparing for an IPO in Hong Kong targeting a massive $50 billion valuation. Such a regulatory investigation could dampen its financial ambitions and undermine its credibility in international markets.
The irony of the situation is not lost on observers: while China is taking these accusations very seriously, it is also doing so to regain control over the narrative pushed by Anthropic. By opening its own investigations, Beijing is ensuring it keeps a grip on the technological story while probing the true extent of its reliance on Western models. As Chinese authorities question the executives and employees of the targeted companies, the sector is watching this confrontation closely, illustrating the limits of digital sovereignty in an AI race where technical boundaries are becoming increasingly porous.