Crypto news

A new scandal is brewing in the artificial intelligence industry. Anthropic, the developer of the advanced Claude model, has uncovered a large-scale campaign of unauthorized distillation of its technologies, orchestrated by operators linked to Chinese giant Alibaba and its AI lab Qwen. This incident has been called the largest of its kind known to the company.

The scale of the attack is impressive: from April 22 to June 5, attackers used nearly 25,000 fake accounts to generate over 28.8 million interactions with the Claude model. According to Anthropic, the goal was to extract Claude's key capabilities in agentic tasks, software development, and long-term planning. Such distillation allows replicating the behavior of an advanced model without the enormous costs of training it.

In a letter addressed to the Chairman of the U.S. Senate Banking Committee, Anthropic emphasizes the audacity and cynicism of the campaign. The company notes that Alibaba, as a public company listed on the New York Stock Exchange and doing business in the U.S., must be accountable to American investors and regulators. According to the developer, such actions not only violate rules but undermine the very economic foundation of U.S. leadership in AI, turning multi-billion dollar American investments into a subsidy for foreign competitors.

Anthropic also warns that unauthorized distillation could accelerate the development of Chinese AI systems used in cyber operations, military tasks, and intelligence. In this regard, the company has appealed to the U.S. Congress with a series of demands. These include expanding technical data sharing between AI developers and the government, clarifying antitrust laws to facilitate information exchange about attacks, and tightening export controls on advanced AI chips and computing resources.

This is not the first time Anthropic has faced mass distillation. Earlier this year, the company already accused Chinese labs DeepSeek, Moonshot AI, and MiniMax of similar actions, which collectively generated millions of interactions with Claude. These incidents highlight the growing tension in the global AI race, where intellectual property and technological leadership are becoming a battlefield. Against this backdrop, Congressman Bill Huizenga has already introduced a bill aimed at combating the extraction of characteristics from closed U.S. AI models by foreign adversaries.

Expert opinion: This incident is a clear signal that the "model war" is entering a new, more aggressive phase. Distillation, long considered a legitimate optimization tool, is now being used as an instrument of industrial espionage on a global scale. The U.S. authorities' response and further steps to protect intellectual property will determine whether the West can maintain its technological advantage or whether we will witness a rapid convergence of capabilities among the world's leading AI systems.