"Spies Target $100M AI Secrets in Code, Says Anthropic CEO"

# China

Anthropic's CEO, Dario Amodei, is seriously concerned about spies, probably from China, snagging valuable "algorithmic secrets" from leading U.S. AI companies. He's calling on the U.S. government to step up and help out. During a chat at the Council on Foreign Relations event on Monday, Amodei pointed out that China's got a rep for "large-scale industrial espionage," and he's pretty sure that AI outfits like Anthropic are in their crosshairs. "Some of these algorithmic secrets are worth a cool $100 million and can be just a few lines of code," he said. "You can bet there are folks out there trying to swipe them, and they might just be pulling it off." Amodei stressed that getting more support from the U.S. government to fend off this threat is "very important," though he didn't spell out exactly what kind of help he's looking for. When TechCrunch reached out to Anthropic for comment on these remarks, they didn't bite but pointed to their recent recommendations to the White House's Office of Science and Technology Policy (OSTP). In their submission, Anthropic pushed for the feds to team up with AI industry bigwigs to boost security at cutting-edge AI labs. This includes working hand-in-hand with U.S. intelligence agencies and their buddies. Amodei's comments fit with his tougher stance on Chinese AI development. He's been pushing for tight U.S. export controls on AI chips to China and slammed DeepSeek for bombing a critical bioweapons data safety test that Anthropic ran. In his essay "Machines of Loving Grace" and other writings, Amodei's worries focus on China using AI for authoritarian and military ends. This kind of position has drawn flak from some in the AI community. They argue that the U.S. and China should be working together more, not less, on AI to avoid an arms race that could lead to one country building a system so powerful that it's beyond human control.

Related article

Comments (36)