Anthropic Accuses Chinese AI Labs of Mass Account Fraud

Anthropic is fighting what it describes as a coordinated campaign by Chinese AI companies to illegally copy its Claude models through mass account fraud and dark web credential theft.

Jacob Klein, Anthropic's head of threat intelligence, told CNBC that foreign adversaries are creating hundreds of thousands of fraudulent accounts to access Claude at scale. The accounts, often purchased through dark web marketplaces selling stolen credit card information, allow competitors to query the models repeatedly and use the responses to train their own systems—a process known as distillation.

The Moonshot allegations

Anthropic has specifically called out Chinese AI lab Moonshot AI, claiming its Kimi K3 model—which gained significant traction in Silicon Valley after launching in July—was illegally trained on the newest version of Claude. The company has also accused DeepSeek, MiniMax, and Alibaba's Qwen team of conducting similar distillation attacks.

"There's an entire illicit ecosystem to try to gain access to Claude and other models," Klein said. "This ecosystem goes through any means necessary to evade our controls, so they can spin up accounts at extreme scale."

The companies named by Anthropic did not respond to requests for comment, according to the report.

Why it matters

The allegations arrive as Anthropic approaches a potential October IPO at a valuation near $1 trillion and as Washington debates how to regulate AI model distillation. The Trump administration signaled in an April memo that distillation undermining American research is "unacceptable" and promised measures to hold foreign actors accountable. The dispute highlights a fundamental tension in AI development: whether distillation represents legitimate competition or intellectual property theft, particularly when conducted across geopolitical boundaries with different legal frameworks and enforcement mechanisms.

Detection challenges

Klein said a telltale sign of illicit distillation is users asking thousands of questions rather than dozens, often across thousands of accounts in a whack-a-mole pattern. The threat extends beyond China to countries like Iran, Russia, and North Korea, where AI companies restrict access due to sanctions.

Travis Lanham, technology chief at cybersecurity firm Armadin and a former Google engineer, noted that bad actors often evade detection because AI companies prioritize accessibility while racing competitors. "These companies are serving billions of requests," Lanham said. "The millions are relatively small compared to everything and it's just sneaking in and trying to look like the rest of the crowd."

Klein emphasized that Anthropic welcomes legal competition and distillation conducted with proper permissions. The concern centers on models created through fraud that lack safety guardrails, potentially enabling surveillance or biological weapons development.

"There is a national security concern at play if malicious actors, bad actors who we don't trust are gaining access to more capable models than they could have otherwise through the act of distillation," Klein said.

OpenAI and Google have published similar reports claiming they face the same distillation threats.

These details were first reported by CNBC.