The general director of Anthropik, Dario Amodea, is frightened that spies, probably from China, gain expensive “algorithmic secrets” from the best AI firms in the USA – and wants the US government to enter.
Speaking in the Council for Foreign Relations event On Monday, Amodei said that China is known for “large -scale industrial espionage” and that AI firms corresponding to anthropic are almost definitely targeted.
“Many of these algorithmic secrets, there are 100 million dollars of secrets, which are a few lines of the code,” he said. “And you know, I’m sure there are people who are trying to steal them and can succeed.”
More help from the US government in defense against this risk is “very important”, added Amodea, without determining exactly what kind of assistance can be required.
Anthropic refused to comment on TechCrunch about comments, but mentioned Anthropic recommendations to the Bureau of Policy of Science and Technology of the White House (OSPP) at the starting of this month.
In the permit, Anthropic (*100*) that the Federal Government should cooperate with the leaders of the AI industry to extend security at the AI Labs Frontier, including through cooperation with American intelligence agencies and their allies.
Notes are in line with the more critical position of Amodea towards the development of Chinese AI. Amodei has called Strong US side export control of AI systems to China, saying that Deepseek won the “worst” in a critical data security test Bioweapons, which carried out anthropic.
Amodea’s fears, as he appointed in his essay “Machines loving grace“And elsewhere, the center of China using artificial intelligence for authoritarian and military purposes.
This kind of attitude led criticism From some in the AI community, which argue that the US and China should cooperate more, no less, in AI, to avoid the arms race, which implies that each country builds a system so powerful that people cannot control it.