Rep. Ro Khanna Requests AI Labs to Report Chinese Cyber Theft Attempts

Representative Ro Khanna has formally requested leading AI labs, including OpenAI, Anthropic, Google, Meta, and SpaceXAI, to report any known attempts by China or other adversaries to illegally access or steal their AI model weights. This inquiry, initiated on October 1, 2026, aims to assess the cybersecurity measures in place to protect these critical assets.

Model weights are essential to AI systems, as they define the parameters of trained models. Khanna warned that a successful theft could severely undermine U.S. leadership in artificial intelligence, stating that such theft could occur with minimal effort.

In addition to outright theft, Khanna raised concerns about the practice of distillation, where new models are trained using the outputs of existing systems rather than directly stealing their weights. He cited potential exploitation by Chinese companies like Moonshot AI and DeepSeek as a significant risk.

This inquiry follows Khanna's previous efforts, including a hearing in April 2026 that examined China's tactics for acquiring AI capabilities. He has also communicated with Chinese AI labs to stress the importance of safeguards against superintelligence risks.

The responses from the targeted companies and any potential disclosures regarding attempted intrusions could have implications for national security and the competitive landscape of the AI industry.