Chinese Firms Used Claude AI for Model Improvements, Anthropic Claims

Three Chinese companies have been accused by Anthropic, the creator of the AI chatbot Claude, of improperly using the technology to enhance their own AI models. The companies, identified as DeepSeek, Moonshot, and MiniMax, reportedly generated over 16 million interactions with Claude using approximately 24,000 fabricated accounts. This activity violates Anthropic's terms of service and regional access restrictions.

The method employed, known as 'distillation,' involves an advanced AI model like Claude evaluating the output of a newer model. This process allows the newer model to learn from the established model's capabilities.

Anthropic warns that such campaigns are escalating in sophistication and pose a threat beyond individual companies or regions.