Congressman Khanna asks AI giants for data on China’s efforts to steal secret code

A Democratic lawmaker warns that Beijing could steal the “weights” of artificial intelligence models developed by OpenAI, Anthropic and other top-tier American companies — erasing the U.S. edge over China in the AI race and posing a risk to American national security, as Reuters reports.

OpenAI and Anthropic have reported numerous cases of Chinese AI companies such as Moonshot and DeepSeek “distilling” their models, that is, training their own tools on the output of leading Western models. But, as Reuters notes, known cases of theft of model weights — the exact parameters and code that make a model work — are few.

Congressman Ro Khanna, who represents Silicon Valley and serves as the ranking member on the House Select Committee on China, has asked the leading American AI companies to provide data on all known efforts by China or other hostile actors to illegally access their prized model weights. He also requested information on the cybersecurity measures in place at these companies to combat such theft.

“The theft of an advanced model weight by a hostile non-state actor could endanger all of humanity, and the theft of such a weight by (China) could erode America’s AI supremacy with a keystroke,” Khanna writes in letters to the chief executives of OpenAI, Anthropic, Google, Meta and SpaceXAI, which Reuters has seen and which had not been previously reported. “Never before have a nation’s national security and economic prospects been so dependent on the cybersecurity of so few companies,” he adds in the letters, dated Wednesday.

The letters illustrate the deepening concern over the national security risks posed by artificial intelligence and add, according to Reuters, to the growing pressure on American AI giants — and on the Trump administration — to rein in the technology. In an interview about the letters, Khanna told Reuters: “I know these guys. You can’t trust Silicon Valley tech billionaires to write the rules that keep us safe.”