Congressman Demanded AI Firm Data on Chinese Access
Representative Ro Khanna has initiated a formal inquiry into model weight security across five major US tech companies.
Updated on Oct. 1, 2026 in Artificial Intelligence
Live Poll
Should the government impose mandatory cybersecurity regulations on private AI companies to protect national security?
Representative Ro Khanna, ranking member of the House China Select Committee, sent letters to the CEOs of OpenAI, Anthropic, Google, Meta, and SpaceXAI requesting data on potential access to model weights by Chinese actors. This inquiry follows reports from OpenAI and Anthropic regarding instances where Chinese firms distilled AI models.
Why it matters
The security of model weights—the foundational parameters learned by an AI during training—is critical to maintaining the American technological lead in artificial intelligence. Khanna’s inquiry aims to assess cybersecurity measures as policymakers debate the risks posed by foreign entities replicating proprietary model capabilities.
Model distillation—a process where a smaller AI model learns to mimic the performance of a larger, more complex model—is the primary mechanism currently under investigation for weight theft. Policymakers are attempting to quantify the extent to which hostile actors have utilized this method to extract proprietary architectures.
The players
Ro Khanna
Representative for Silicon Valley and ranking member of the House China Select Committee focusing on tech industry oversight.
OpenAI
An AI research and deployment company known for its large-scale language models and proprietary API infrastructure.
Anthropic
An AI research firm focused on safety and large-scale model development that reported instances of model distillation.
President Trump
The current President of the United States overseeing national policy on artificial intelligence and industry standards.
The details
Model weights represent the essential numerical values stored within a neural network that dictate how it processes information. By distilling these weights, external actors can theoretically recreate the performance of a proprietary model without having access to the original training infrastructure or the vast datasets required for the initial development. Khanna is now seeking data on how firms protect these weights from unauthorized access, following voluntary AI standards discussions recently referenced by the President.
Timeline
September 30, 2026: Representative Ro Khanna dated letters to the CEOs of five major AI companies.
September 30, 2026: President Trump noted that technology executives have agreed to voluntary AI safety standards.
The Tech Race
This inquiry follows the implementation of voluntary AI standards recently discussed by the President. It marks a shift from industry-led self-regulation toward more direct congressional oversight regarding the protection of strategic intellectual property.
This inquiry represents an early stage of political scrutiny that may eventually lead to new regulatory mandates for how AI companies store and protect their source code. For developers and industry participants, this signals a potential increase in compliance requirements regarding data security and cross-border research collaboration.
The takeaway
The trajectory of this investigation suggests that model weight protection will remain a primary focus for congressional committees managing national security risks. Stakeholders should monitor future disclosures from the House China Select Committee regarding the effectiveness of currently implemented cybersecurity measures.
Further reading
For broader context on the industry-wide security challenges, explore the latest updates from our Artificial Intelligence desk.
Live Poll
Should the government impose mandatory cybersecurity regulations on private AI companies to protect national security?
