The AI model Claude has been at the center of a high-stakes cybersecurity battle, with Chinese tech giant Alibaba accused of launching the largest attack yet to clone its capabilities. Anthropic, the developer behind Claude, claims that Alibaba's operators used thousands of fake accounts to bypass safety barriers and extract answers from Claude's most valuable capabilities.

The alleged attacks occurred between April 22 and June 5, with over 28.8 million exchanges with Claude through almost 25,000 fraudulent accounts. Anthropic warned that these distillation attacks will help China reach Mythos Preview-level capabilities sooner if left unchecked.

Background and Context

The development of advanced AI models like Claude has sparked a global competition for technological supremacy. China's efforts to match the capabilities of US frontier models have been ongoing, with Anthropic's Mythos model being a particular target. In April, Donald Trump accused China of "industrial-scale" AI theft after Anthropic accused Chinese firms DeepSeek, Moonshot, and MiniMax of using similar tactics.

Anthropic's CEO, Dario Amodei, has been vocal about the need for stronger regulations to prevent these types of attacks. In a letter sent to US Senators Tim Scott and Elizabeth Warren, Anthropic shared "new, confidential evidence of the largest campaign to illicitly extract Claude's capabilities we have ever measured."

Why it Matters to the Industry

The implications of these distillation attacks are far-reaching, with Anthropic warning that they will help China reach Mythos Preview-level capabilities sooner. This could give Chinese firms an unfair advantage in the global AI market, allowing them to harvest and repackage US AI capabilities at a fraction of the cost.

Anthropic also cited the US Department of Defense's claims that Alibaba and several major firms like car maker BYD and tech company Baidu are tied to the Chinese military. While these allegations have been denied by the companies, they highlight the growing concern about the potential misuse of AI technology for national security purposes.

What Comes Next

Anthropic is urging Congress to pass legislation with three objectives: updating antitrust laws to allow AI firms to share information about evolving Chinese tactics, imposing export controls on chips to hamstring Chinese access to advanced compute, and penalizing Chinese labs' "bad behavior" so that it's more difficult and costly to rely on distillation attacks.

The company also recommended that Congress pass laws limiting Chinese firms from accessing US models or advanced US chips or from relying on data centers outside of China. These measures aim to prevent the misuse of AI technology for national security purposes and maintain American leadership in the global AI market.

Key Facts

  • Anthropic accused Alibaba of launching the largest attack yet to clone Claude's capabilities, with over 28.8 million exchanges through almost 25,000 fraudulent accounts.
  • The alleged attacks occurred between April 22 and June 5, with Anthropic warning that they will help China reach Mythos Preview-level capabilities sooner if left unchecked.
  • Anthropic shared "new, confidential evidence of the largest campaign to illicitly extract Claude's capabilities we have ever measured" in a letter sent to US Senators Tim Scott and Elizabeth Warren.
  • The company urged Congress to pass legislation with three objectives: updating antitrust laws, imposing export controls on chips, and penalizing Chinese labs' "bad behavior".
  • Anthropic also recommended that Congress pass laws limiting Chinese firms from accessing US models or advanced US chips or from relying on data centers outside of China.
  • The alleged attacks highlight the growing concern about the potential misuse of AI technology for national security purposes, with Anthropic warning that distillation attacks will help China reach Mythos Preview-level capabilities sooner if left unchecked.