US AI Giant Anthropic Accuses Chinese Rivals of Mass Data Theft
In a significant development, the US artificial intelligence company Anthropic has publicly accused three Chinese AI firms of engaging in industrial-scale intellectual property theft. The allegations, announced on Monday, center on the illicit extraction of capabilities from Anthropic's Claude chatbot using a technique known as "distillation." This method involves using outputs from a more powerful AI system to rapidly enhance the performance of a less capable one, effectively siphoning off advanced features without independent development.
Growing Intensity and Sophistication of Theft Campaigns
Anthropic stated that the campaigns by DeepSeek, Moonshot AI, and MiniMax are "growing in intensity and sophistication," with the company warning that "the window to act is narrow." According to Anthropic, these firms achieved their ends through approximately 16 million exchanges with its Claude model and 24,000 fake accounts. This allowed them to bypass export controls on powerful US technology, which are designed to maintain American dominance in the sensitive AI sector.
The practice of distillation is not new in AI development; it is commonly used by companies to create cheaper, smaller versions of their own models. However, it gained widespread attention a year ago when DeepSeek released a low-cost generative AI model that performed at a level comparable to ChatGPT and other top American chatbots, challenging assumptions of US superiority.
National Security Risks and Industry Response
Anthropic has raised serious national security concerns, arguing that models built through illicit distillation are unlikely to retain safety guardrails. These guardrails are crucial for preventing misuse, such as restrictions on helping develop bioweapons or enabling cyberattacks. The company emphasized that this issue poses risks that no single company can tackle alone, calling for coordinated industry and government responses.
Anthropic's arch-rival, OpenAI, creator of ChatGPT, made similar accusations to US lawmakers earlier this month. OpenAI claimed that Chinese companies are using distillation techniques as part of "ongoing efforts to free-ride on the capabilities developed by OpenAI and other US frontier labs." This highlights a broader pattern of alleged intellectual property theft in the AI industry.
Details of the Alleged Operations
Among the accused firms, MiniMax reportedly ran the largest operation, generating more than 13 million exchanges. The campaigns focused heavily on areas where Claude is considered a leader, such as coding, agentic reasoning, and tool use. To circumvent Anthropic's ban on commercial access from China, the labs allegedly routed traffic through proxy services that managed vast networks of fraudulent accounts.
This case underscores the escalating tensions in the global AI race, with US companies increasingly vocal about protecting their technological advancements. As the industry grapples with these challenges, the call for stronger safeguards and international cooperation becomes more urgent to address the sophisticated threats posed by such theft campaigns.
