Leading Democrat asks AI firms for data on any Chinese access to sensitive code

• OpenAI and Anthropic have reported multiple instances of Chinese AI companies like Moonshot and DeepSeek training their own tools using output from leading Western models through a process called 'distilling'
• Khanna warned that theft of advanced model weights by China 'could erode America's AI lead with the stroke of a keyboard' and requested information on cybersecurity measures protecting these assets
• The letters highlight tensions between calls for AI regulation from lawmakers and President Trump's focus on maintaining competitive advantage over China, with Khanna stating 'You cannot trust Silicon Valley tech billionaires to write the rules to keep us safe'

Summary

Democratic Congressman Ro Khanna, ranking member of the House China Select Committee, has requested data from five major U.S. AI companies—OpenAI, Anthropic, Google, Meta, and SpaceXAI—regarding Chinese attempts to illegally access sensitive AI model weights. The letters, dated Wednesday and seen by Reuters, warn that theft of advanced model weights could eliminate America's AI advantage and threaten national security.

Key Developments:

• OpenAI and Anthropic have reported multiple instances of Chinese AI firms, including Moonshot and DeepSeek, "distilling" their AI models by training tools from Western model outputs
• While distillation cases exist, few confirmed instances of actual model weight theft have been documented
• Model weights are the exact parameters and code that make AI systems operate, representing the most valuable intellectual property

National Security Concerns:

Khanna emphasized that "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 warned that theft by hostile actors "could endanger all of humanity."

Political Context:

The inquiry adds pressure on AI companies and the Trump administration to strengthen oversight. President Trump has downgraded such concerns, prioritizing U.S. competitive advantage over China and pointing to existing Justice Department safeguards. Trump recently announced tech executives agreed to establish voluntary standards for AI.

Khanna stated skepticism about industry self-regulation: "You cannot trust Silicon Valley tech billionaires to write the rules to keep us safe."

The companies and Chinese Embassy did not immediately respond to requests for comment.