Anthropic CEO says spies are after $100M AI secrets in a ‘few lines of code’

Dario Amodei from Anthropic is concerned that spies, probably from China, is given costly “algorithmic secrets” from the top AI companies in the USA -and he wants the US government to enter.

Speak in a council for foreign relationships Event On Monday, Amodei said that China was known for his “large-scale industrial espionage” and that AI companies such as Anthrope are almost certainly targeted.

“Many of these algorithmic secrets, there are $ 100 million secrets that are a few code lines,” he said. “And I am sure that there are people who try to steal them and they can be successful.”

More help from the US government to defend itself against this risk was “very important”, Add, Add, without exactly indicating what kind of help would be required.

Anthropic rejected it to comment on the comments, but which refers to Recommendations of Anthropic At the beginning of this month to the Office for Science and Technology Policy (Ostp) of the White House.

In the submission, Anthropic argues that the Federal Government should work with the executives of the AI ​​industry to improve the security in Frontier AI Labs, also by working with US secret services and its allies.

The remarks correspond to the more critical attitude of Amodei on Chinese AI development. Amodei has prompt Strong US export controls for AI chips to China While he said Deepseek scored “the worst” In a critical bioweapons data security test that was carried out anthropically.

Amodeis consider when he explained in his essay “Machines of loving graceAnd elsewhere they concentrate on China to use AI for authoritarian and military purposes.

This type of attitude has led to this criticism From some in the AI ​​community that argue that the USA and China should work together more and no less on the AI ​​to avoid arms arms that builds up a system in both countries that is so powerful that people cannot control it.