American AI giant Anthropic has officially accused Chinese research labs, including DeepSeek, of mass data scraping from its Claude model to train their own systems. Simultaneously, the company faces internal pressure: U.S. Secretary of Defense Pete Hegseth has summoned Anthropic's CEO for an urgent meeting regarding restrictions on the military use of their technology, while Elon Musk's competing xAI has already secured a deal with the Pentagon.

Espionage Allegations

Anthropic claims Chinese firms used thousands of accounts to steal the structure of the Claude model.

Pentagon's Ultimatum

Secretary of Defense Pete Hegseth demands the removal of barriers to the military use of artificial intelligence.

Contract for xAI

Elon Musk's company signed a deal to deliver the Grok model to the army's classified systems.

Violation of Chip Sanctions

There are suspicions that DeepSeek trained its models on banned Nvidia H100 chips.

The American artificial intelligence sector has become an arena for an international conflict of espionage and military character. Anthropic revealed that three Chinese entities – DeepSeek, Moonshot AI, and MiniMax – used over 24,000 fake accounts to mass-download data from the Claude model. This process, called "distillation," involves using the responses of a high-class model to train cheaper and smaller systems, which Anthropic describes as intellectual property theft on an unprecedented scale. Findings also suggest that DeepSeek may have violated U.S. sanctions by training its algorithms on the latest Nvidia processors, whose export to China is theoretically banned. Concurrently, a conflict is escalating between Anthropic and the Donald Trump administration. Secretary of Defense Pete Hegseth summoned the company's CEO, Dario Amodei, to a meeting at the Pentagon. The reason is Anthropic's refusal to allow fully unrestricted use of its technology in combat systems, which the company argues is due to safety and ethical concerns. The Pentagon has threatened to exclude the company from future government contracts if these restrictions are not lifted. The situation is worsened by the fact that Elon Musk's startup, xAI, has just announced an agreement with the military to implement the Grok model into classified defense systems. The technological rivalry between the USA and China, referred to as a "digital cold war," intensified after 2022 when Washington introduced drastic restrictions on semiconductor trade. Historically, technology transfer and reverse engineering have been key elements of Beijing's modernization strategy in the high-tech sector. „This is not healthy competition; it is the systematic mining of our hard work by state-aligned actors.” — Dario Amodei Analysts indicate that the Pentagon's pressure on tech companies could lead to the creation of a new military doctrine based on autonomous lethal systems. While Anthropic tries to maintain control over the ethical use of AI, the market and political successes of xAI put the company in a difficult negotiating position. If Anthropic does not yield to Hegseth's demands, it risks not only losing lucrative contracts but also marginalization in the U.S. defense ecosystem, which is crucial for national security. Number of Accounts Used for Data Download: DeepSeek: 12, MiniMax: 7, Moonshot AI: 5

Mentioned People

  • Dario Amodei — CEO and co-founder of Anthropic, summoned by the Pentagon for explanations.
  • Pete Hegseth — U.S. Secretary of Defense in the Trump administration, proponent of military use of AI.
  • Elon Musk — Owner of xAI, who is strengthening cooperation with the U.S. military.