Researchers find Chinese hacking campaigns targeting AI firms, Asian governments

A Democratic lawmaker has urged major US AI companies to disclose any incidents of Chinese access to their sensitive model weights. Congressman Ro Khanna, ranking member on the House China Select Committee, has requested information from OpenAI, Anthropic, Google, Meta, and SpaceXAI regarding any unauthorized attempts by China or other hostile actors to steal these crucial AI components.

He also inquired about the cybersecurity measures employed by these firms to prevent such thefts. Khanna's concern is that the theft of an advanced model weight by a hostile non-state actor could pose a significant threat to global security, potentially eroding the US AI advantage. He stressed that this issue is of utmost importance, as the nation's economic prospects and national security are increasingly reliant on the cybersecurity of just a few AI companies.

The Chinese Embassy in Washington and the aforementioned companies have not yet responded to requests for comment. The letters from Khanna demonstrate growing apprehension about the national security risks associated with AI, and they reflect mounting pressure on American AI titans and the Trump administration to regulate the technology more closely.

Both major parties have expressed calls for action on this issue, but President Trump has downplayed these concerns, emphasizing the need to maintain America's competitive advantage over China and citing the Justice Department as an existing safeguard for AI. Khanna, however, maintains that Silicon Valley tech billionaires are not capable of writing effective safety rules for AI.