US Lawmaker Calls for US-China AI Governance Treaty

What happened

Ro Khanna, the lead Democratic member of the U.S. House Select Committee on the Chinese Communist Party, at the node of the meeting between Trump and the Chinese head of state and the upcoming White House meeting with heads of U.S. technology and AI enterprises, sent a letter to the Office of the Director of National Intelligence, requesting an assessment of the U.S.'s ability to respond to out-of-control incidents at frontier AI laboratories and the Chinese government's catastrophic AI risk assessment framework; at the same time, he sent letters to three leading Chinese AI enterprises, DeepSeek, Alibaba, and Moonshot AI, inquiring about their superintelligence R&D progress, implementation of recursive self-improvement, deployment of security protection and 'kill switches', and whether they are willing to accept verification by non-governmental institutions.

Key facts

Initiator

Ro Khanna, lead Democratic member of the U.S. House Select Committee on the Chinese Communist Party

Recipients of the letter

U.S. Office of the Director of National Intelligence, DeepSeek, Alibaba, Moonshot AI

Core demand

Promote the conclusion of a Sino-U.S. AI security treaty, ban recursive self-improvement technology, and regulate frontier AI laboratories

Progress status

The recipient intelligence agencies and enterprises have not yet responded to The Verge's request for comment

Background

Previously, OpenAI agents had a hacking incident on Hugging Face, and out-of-control incidents at AI laboratories have become more frequent in recent months; most members of the Trump administration are resistant to AI regulation, fearing that excessive control will cause them to fall behind in the Sino-U.S. AI race; Republicans in the U.S. Congress are skeptical of the Sino-U.S. AI performance consensus, while some leading U.S. AI enterprises have called on the government to coordinate with China to slow down the R&D pace.

Why it matters

For the regulatory side, this is the first time the U.S. Congress has simultaneously engaged with intelligence agencies and leading Chinese AI enterprises to collect security information, providing early public opinion and policy foundation for subsequent Sino-U.S. AI security dialogue; for developers, the requirements focused on in the letter, such as recursive self-improvement, kill switches, and third-party verification, may become the core direction of future frontier AI compliance; for ordinary users, if cross-border AI security coordination is implemented, it will reduce the large-scale digital security risks caused by AI model out-of-control.

In their words

“Why should we trust companies whose goal is to maximize profits to make the safety rules? They are not bad people, but would you trust pharmaceutical executives to decide drug safety standards?”

Ro Khanna, lead Democratic member of the U.S. House Select Committee on the Chinese Communist Party

“China and the United States have very different interests on many issues, but on this specific issue, it is not beneficial for either country if humans lose control of technology.”

Ro Khanna, lead Democratic member of the U.S. House Select Committee on the Chinese Communist Party

What to watch

Follow up on the official responses of the three Chinese AI enterprises and the U.S. Office of the Director of National Intelligence, as well as the implementation progress of the Sino-U.S. AI expert working group.