Anthropic, a company specializing in artificial intelligence, has raised serious allegations against three Chinese AI firms—DeepSeek, Moonshot, and MiniMax—for engaging in unauthorized "distillation" attacks. These firms reportedly created 24,000 accounts and conducted 16 million interactions with Anthropic's large language model, Claude, to illegally enhance their own AI systems.
In a blog post released on Sunday, Anthropic detailed that these distillation attacks involved using a less capable AI model to learn from the outputs of Claude, a more advanced model. While distillation is a common and legitimate practice within AI development, Anthropic highlighted that it can also be misused by competitors to rapidly and cheaply acquire sophisticated AI capabilities.
The attacks allegedly targeted Claude's unique features such as agentic reasoning, coding, data analysis, and computer vision. "Each campaign focused on Claude's most distinct capabilities: agentic reasoning, tool use, and coding," stated the company, which is valued at several billion dollars.
Anthropic identified the perpetrators through a combination of IP address analysis, request metadata, infrastructure indicators, and corroborations from industry partners who noticed similar activities on their platforms. DeepSeek, Moonshot, and MiniMax—all headquartered in China—are high-profile AI companies, with DeepSeek being the most internationally recognized.
Beyond the breach of intellectual property, Anthropic warned of the broader geopolitical risks these distillation campaigns pose. The firm expressed concerns that foreign AI models distilled from U.S. technology could be incorporated into military, intelligence, and surveillance systems, empowering authoritarian regimes to execute cyberattacks, spread disinformation, and conduct extensive surveillance.
In response, Anthropic plans to bolster its defenses by improving detection systems, sharing threat intelligence, and tightening access controls. The company also urged for increased collaboration among domestic AI firms and policymakers to thwart such foreign threats. "No single company can counter this alone. A coordinated industry-wide response is essential," Anthropic emphasized.
Anthropic's call for action is a significant move towards safeguarding AI innovations from unauthorized exploitation, highlighting the need for collective industry efforts to protect intellectual property and national security.