Why Anthropic is right to fear Chinese AI distillation

Anthropic, the world's leading AI company at this point, arguably, has accused the Chinese tech giant Alibaba of illicitly training its AI models on the output... of Anthropix AI model, Claude.

It claims that Alibaba set up around 25,000 fake user accounts for Claude so that it could get the outputs from Anthropix model and use them to make its own AI model smarter.

And this is a process called distillation.

Anthropic laid out its accusations in a letter to a couple of US senators, Tim Scott and Elizabeth Warren.

And according to the Wall Street Journal, which got a copy of the letter, it said, "...Alibaba's campaign targeted some of Claude's most valuable capabilities, such as agentic reasoning, software engineering, and long-horizon tasks." It claimed that this was turning, quote, billions of dollars in American investment and R&D into a massive subsidy for our geopolitical competitors.

Does Anthropic actually have a leg to stand on here? I'll get to the details in a moment, but the short answer is no.

And that is why the company is trying to get the American government on board here.

It said in a statement given to the Wall Street Journal, we believe combating the threat of illicit distillation requires coordinated action between government and industry, and we will continue working with Congress and the administration to maintain American AI leadership.