What exactly is adversarial distillation, and why is it a problem?

Adversarial distillation is a method where developers train new AI systems using results from another company's cutting-edge AI model to replicate its capabilities at a fraction of the cost. Anthropic warns that systems built this way often lack safety guardrails and that the practice violates AI labs' terms of use when employed without permission to replicate leading models.

How large was Alibaba's alleged access campaign?

According to Anthropic's letter, the campaign involved almost 25,000 fraudulent accounts that generated 28.8 million exchanges with Claude between April and June.

How are U.S. AI companies responding to this threat?

Anthropic, OpenAI, and Google have joined forces to share information about distillation attempts that violate their terms of service. Anthropic has also sent a letter to U.S. senators and White House officials urging the Trump administration to halt the practice.

Back to articles

Anthropic accuses Alibaba of using thousands of fraudulent accounts to illicitly access its Claude AI model, marking what it calls the largest attempt by a Chinese company to copy U.S. AI technology.

Japan Times Tech22h ago5 min read

Key takeaway

Anthropic has accused Alibaba of orchestrating a massive campaign to illegally access its Claude AI model through thousands of fake accounts between April and June. The U.S. AI developer argues that Chinese companies are systematically copying advanced AI systems using a low-cost technique that strips away safety protections, and has called on U.S. officials to crack down on the practice.

Summaries like this, in your inbox every morning.

3 Key Points

What happened
Anthropic alleges that operators linked to Alibaba's Qwen AI lab conducted a campaign involving almost 25,000 fraudulent accounts that generated 28.8 million exchanges with Claude between April and June, targeting the model's software engineering and reasoning capabilities.
Why it matters
Anthropic warns that Chinese labs are systematically harvesting U.S. AI capabilities through a practice called adversarial distillation—training new AI systems using results from leading U.S. models to build rival chatbots at far lower cost—and that systems built this way often lack safety guardrails. The company has urged the Trump administration to step up enforcement.
What to watch
Anthropic, OpenAI, and Google have formed a coalition to share information about distillation attempts that violate their terms of service, signaling industry-wide concern about the practice.

FAQ

What exactly is adversarial distillation, and why is it a problem?: Adversarial distillation is a method where developers train new AI systems using results from another company's cutting-edge AI model to replicate its capabilities at a fraction of the cost. Anthropic warns that systems built this way often lack safety guardrails and that the practice violates AI labs' terms of use when employed without permission to replicate leading models.
How large was Alibaba's alleged access campaign?: According to Anthropic's letter, the campaign involved almost 25,000 fraudulent accounts that generated 28.8 million exchanges with Claude between April and June.
How are U.S. AI companies responding to this threat?: Anthropic, OpenAI, and Google have joined forces to share information about distillation attempts that violate their terms of service. Anthropic has also sent a letter to U.S. senators and White House officials urging the Trump administration to halt the practice.

Discussion

No discussion yet for this article

Stay ahead with AI news

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Get Started Free

Free · takes 30 seconds · unsubscribe anytime

5 minutes a day. The AI essentials.

200+ sources · Email / LINE / Slack

Get it free →