U.S. AI startup Anthropic expands restrictions on Chinese entities

U.S. artificial intelligence (AI) company Anthropic has announced the expansion of restrictions on Chinese entities to include subsidiary companies and organizations based overseas, citing national security risks.

Companies based in China, as well as in other countries, including Russia, North Korea and Iran, are already unable to access Anthropic's commercial services "due to legal, regulatory, and security risks," the company said in a statement dated September 5.

Yet some groups "continue accessing our services in various ways, such as through subsidiaries incorporated in other countries," it said.

Anthropic said that the change would affect entities that are more than 50 percent owned, directly or indirectly, by companies in unsupported regions.

The update to the company's terms of service affects such entities "regardless of where they operate," it said.

"This is the first time a major U.S. AI company has imposed a formal, public prohibition of this kind," said Nicholas Cook, a lawyer focused on the AI industry with 15 years of experience at international law firms in China.

"The immediate commercial effect may be modest, since U.S. AI providers already face barriers to operating in this market and relevant groups have been self-selecting for their own locally developed AI tech," he told AFP.

An Anthropic executive told the Financial Times that the move would have an impact on revenues in the "low hundreds of millions of dollars."

The San Francisco-headquartered company, which is valued at $183 billion and backed by Amazon, is known for its Claude chatbot and AI models. Founded in 2021 by former executives from OpenAI, the company recently announced that it had raised $13 billion in its latest funding round, saying it now has more than 300,000 business customers.

When asked about Anthropic's announcement at a regular briefing, Chinese Foreign Ministry spokesperson Guo Jiakun said he was not familiar with the specifics but stressed that "China opposes politicizing trade and sci-tech issues and using them as a weapon and a tool."

"Such practice does no one good," said Guo.

Anthropic recently agreed to pay a landmark $1.5 billion to settle a class-action lawsuit from a group of authors who accused the company of pirating millions of books to train its Claude chatbot.

Lawyers for the authors called the settlement the largest copyright recovery in history and the first of its kind in the AI era.

(With input from AFP)