Anthropic Accuses Chinese AI Labs of Model Theft

Serdar HocamAuthor & Editor

Anthropic has targeted Chinese labs, stating that Claude models were illegally copied using the dark web and fraudulent methods.

◉ 0 views
Anthropic's distillation battle turns to the dark web as China concerns swell

AI company Anthropic announced that Chinese labs, including Moonshot AI, DeepSeek, MiniMax, and Alibaba, illegally copied Claude models using illicit methods and the dark web.

Illegal Model Copying Accusations

Jacob Klein, head of threat intelligence at Anthropic, stated that activities originating from the Chinese market have reached the level of theft rather than competition. It was stated that foreign actors accessed Claude models to produce competitive technology.

Targeted Chinese AI Companies

Anthropic claimed that the Kimi K3 model released in July was illegally trained using the newest version of Claude. The company also accused Moonshot, DeepSeek, MiniMax, and Alibaba of organizing massive infiltration and copying attacks.

Dark Web and Fake Account Tactics

Cybersecurity experts also drew attention to threats originating from countries such as Iran, Russia, and North Korea. It was reported that malicious labs use dark web markets for stolen credit card information and compromised accounts.

National Security and Espionage Concerns

Jacob Klein emphasized that malicious actors gaining access to advanced models without any guardrails creates national security issues. Experts noted that AI companies struggle to detect these attacks while maintaining platform accessibility.