Anthropic says DeepSeek and other Chinese AI companies fraudulently used Claude (www.businessinsider.com)

Anthropic has accused three major Chinese AI labs—DeepSeek, MiniMax, and Moonshot AI—of illegally using its AI model, Claude, to enhance their own technologies through a method known as distillation. This process involves training less powerful models on the outputs of more powerful ones, a practice that is legitimate in many contexts but is reportedly being exploited by these competitors to gain a significant advantage in the AI arms race. Anthropic claims that these labs orchestrated large-scale operations using around 24,000 fraudulent Claude accounts that generated over 16 million interactions, all in violation of terms and regional restrictions. The implications of these accusations extend beyond corporate competition; they highlight the need for tighter controls on chip exports as a defense against such illicit activities. Anthropic asserts that distillation attacks pose security risks, especially as inadequately trained models could be misused for harmful purposes, including bioweapons development. In response to these threats, Anthropic has implemented behavioral fingerprinting systems and shared intelligence with other AI firms to mitigate risks, while also advocating for coordinated regulatory action. This episode underscores growing concerns within the AI community regarding the ethical and security ramifications of AI technology misuse, insisting on the importance of safeguarding intellectual property in a rapidly evolving digital landscape.