Full Breakdown
Democratic Lawmaker Requests AI Firms’ Data on Chinese Access to Model Weights
By Drooid · · How we work
Congressional Request Highlights AI Security Concerns
On October 1, Congressman Ro Khanna, ranking member of the House China Select Committee, sent letters to the CEOs of OpenAI, Anthropic, Google, Meta and SpaceX AI. In the letters he asked each firm to disclose any known attempts by China—or other hostile actors—to illegally obtain their model weights, the detailed parameters and code that enable AI systems to function. Khanna warned that theft of such weights by a hostile non-state actor could endanger humanity and that a breach by China could erode the United States’ lead in artificial intelligence with “the stroke of a keyboard.”
Context of Chinese AI Model Access
OpenAI and Anthropic have previously reported that Chinese AI companies, including Moonshot and DeepSeek, have “distilled” their models or trained new tools using output from leading Western systems. While these reports confirm instances of model-output copying, the firms note that confirmed cases of direct theft of model weights remain rare. The broader concern stems from a growing body of research suggesting that advanced AI agents could act unpredictably, with some experts warning of potential rogue behavior within a decade.
Responses from Companies and Government
The Chinese Embassy in Washington and the AI firms did not immediately comment on Khanna’s request. OpenAI and Anthropic reiterated their earlier reports of Chinese distillation activities, and Meta and Google indicated they have observed similar behavior by their agents. President Donald Trump, speaking on the same day, downplayed the security alarm, emphasizing the importance of maintaining U.S. competitive advantage over China and noting that the Justice Department already provides a regulatory guardrail for AI. He also referenced voluntary industry standards that tech executives have agreed to develop.
Potential Implications for U.S. National Security
Khanna’s letters underscore mounting congressional pressure to address AI-related national-security risks. By seeking detailed disclosures, the inquiry aims to assess the effectiveness of existing cybersecurity measures at a handful of companies whose technologies are deemed critical to the nation’s economic and security interests. The request adds to bipartisan calls for clearer oversight of AI development and may influence future legislative or regulatory actions targeting the protection of advanced model weights from foreign exploitation.