US Lawmaker Presses OpenAI, Anthropic for Data on Chinese Attempts to Steal AI Code
Congressman Ro Khanna asked five top US AI firms to disclose any efforts by China to access their model weights, warning the theft could erase America's lead.
WASHINGTON, Oct 1 (Reuters) - A Democratic lawmaker is warning that Beijing could steal AI model weights developed by OpenAI, Anthropic and other top US firms, erasing the US edge over China in the AI race and posing a risk to US national security.
OpenAI and Anthropic have reported multiple instances of Chinese AI firms like Moonshot or DeepSeek "distilling" their AI models, or training their own tools from the output of leading Western models. But there are few known cases of any bad actor stealing model weights, the exact parameters and code that make an AI model operate.
Congressman Ro Khanna, who represents Silicon Valley and serves as ranking member on the House China Select Committee, asked top US AI firms 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 cybersecurity measures in place at the AI firms 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 model weight by (China) could erode America's AI lead with the stroke of a keyboard," he wrote in letters to the CEOs of OpenAI, Anthropic, Google , Meta , and SpaceXAI seen by Reuters and not previously reported.
"Never before have the national security and economic prospects of a nation been so dependent on the cybersecurity of a small handful of companies," he added in the letters, dated Wednesday.
The Chinese Embassy in Washington and the companies did not immediately respond to requests for comment.
The letters illustrate deepening concern about the national security risks posed by AI and add to mounting pressure on American AI giants – and the Trump administration – to rein in the technology.
OpenAI and Anthropic have reported that their AI agents hacked into external secure systems, while Meta and Google say they have found similar behavior by their agents. Such breaches have prompted some researchers to warn that the technology could go rogue and kill humans within a decade.
The warnings have fueled calls for action from both Democratic and Republican lawmakers. But President Donald Trump has repeatedly downplayed such concerns, stressing the importance of maintaining America's competitive edge over China and pointing to the Justice Department as a pre-existing guardrail for AI.
Trump on Tuesday said tech executives had agreed to establish voluntary standards for AI.
In an interview about the letters, Khanna said: "I know these guys. You cannot trust Silicon Valley tech billionaires to write the rules to keep us safe."
(Reporting by Alexandra Alper; Editing by Stephen Coates)
