Chinese AI labs accused by Anthropic of mining Claude

Anthropic has alleged that Chinese AI labs DeepSeek, Moonshot, and MiniMax launched separate industrial-scale distillation campaigns aimed at its Claude large language model to fortify their own LLMs, CyberScoop reports.

DeepSeek, Moonshot, and MiniMax had leveraged 24,000 fraudulent accounts to deliver more than 16 million requests to Claude in a bid to extract the LLM's capabilities at scale, according to Anthropic.

"Foreign labs that distill American models can then feed these unprotected capabilities into military, intelligence, and surveillance systems enabling authoritarian governments to deploy frontier AI for offensive cyber operations, disinformation campaigns, and mass surveillance," said Anthropic, which has advocated for more stringent Chinese chip export controls after the exploitation of Claude in a Chinese cyberespionage campaign.

Intellectual property and computing power pilfered in distillation campaigns were noted by Oligo Security co-founder and Chief Technology Officer Gal Elbaz to potentially result in more destructive cyberattacks.