Congressman pushes for US-China AI safety treaty as leaders drag feet

Rep. Ro Khanna (D-CA), the top Democrat on the House Select Committee on China, is making the case that the world's two largest AI powers need to strike a binding agreement on artificial intelligence safety—but he's skeptical either government will move with the urgency the moment demands.

In exclusive letters to U.S. intelligence officials and major Chinese AI firms, Khanna is raising alarms about both countries' preparedness for catastrophic AI incidents and their commitment to controlling the technology's risks. His push comes as President Trump and Xi Jinping recently met at the White House, a conversation Khanna says fell short of addressing AI's existential stakes.

The proposal#

Khanna's ultimate goal is a formal U.S.-China treaty that would ban recursive self-improvement (RSI)—the ability for AI systems to improve themselves without human intervention—and establish monitoring of frontier AI labs. The letters to Chinese firms including DeepSeek, Alibaba, and Moonshot AI request documentation on their plans for achieving superintelligence, their safety safeguards, "kill switches," and willingness to submit to third-party inspections under any future agreement.

He has also written to the Office of the Director of National Intelligence asking it to assess how well the U.S. could respond to an AI lab losing control of its systems—a scenario he says is becoming more common—and to evaluate China's own risk assessment capabilities.

Political obstacles#

Both Trump and House Speaker Mike Johnson have signaled resistance to tighter AI oversight, with Trump posting on Truth Social that he wants to "leave" superintelligence "exactly where it is." Johnson said on Fox Business he expects discussions to focus on balancing oversight without slowing innovation, citing fears of losing the AI race to China.

Republican skeptics, such as House China Committee Chair John Moolenaar, have questioned whether Beijing would honor any treaty. Meanwhile, Chinese companies have previously ignored Khanna's requests to commit to slowing development.

Why trust matters#

Khanna is equally critical of American tech companies' role in safety discussions. "Why should we trust companies that have a motive to maximize profits to be making the rules for what is safe?" he asked . He likened it to letting pharmaceutical executives decide drug safety standards.