Anthropic Accuses 3 Chinese Companies of Data Harvesting

Anthropic Accuses 3 Chinese Companies of Data Harvesting

SAN FRANCISCO—Anthropic, a leading San Francisco-based artificial intelligence startup, has raised serious concerns about three prominent Chinese companies improperly using its AI-generated data to develop their own systems. The company revealed that DeepSeek, Moonshot, and Minimax utilized approximately 24,000 fraudulent accounts to create over 16 million conversations with Anthropic's Claude chatbot.

These practices, according to Anthropic, violate their terms of service and pose significant national security risks. By harvesting data from American AI systems, Chinese companies may gain the capability to develop advanced technologies that could be used for harmful purposes, such as creating bioweapons or enabling mass surveillance.

Anthropic also highlighted its ongoing legal battles, including a $1.5 billion settlement reached last year for unauthorized use of copyrighted data in training its systems. This case underscores the importance of protecting intellectual property and ensuring responsible AI development practices.

The startup has warned that such data harvesting activities are becoming more sophisticated and frequent, emphasizing the need for coordinated action among industry players, policymakers, and the global AI community to address these challenges effectively.