Rep. Ro Khanna, the top Democrat on the House Select Committee on China, is calling for a treaty between the United States and China to manage risks associated with advanced artificial intelligence, specifically targeting recursive self-improvement (RSI). The move follows a recent meeting between President Donald Trump and Chinese President Xi Jinping, which Khanna described as constructive but inadequate regarding the urgency of AI safety.
What Happened
In letters shared exclusively with The Verge, Khanna requested that the Office of the Director of National Intelligence (ODNI) assess how effectively the US might respond to an AI lab losing control of its bots. He also asked the agency to evaluate the Chinese government’s approach to assessing catastrophic AI risks. The goal, according to Khanna, is to gather information that will help Congress and the White House craft an agreement with their Chinese counterparts.
Khanna also sent letters to three major Chinese AI companies—DeepSeek, Alibaba, and Moonshot AI—asking for documentation on their efforts to achieve "superintelligence" and recursive self-improvement. The letters inquired about safeguards, "kill switches," and whether the companies would commit to inspections by a non-governmental agency if a treaty were established. Neither the companies nor the ODNI immediately responded to requests for comment.
Why It Matters
The push for a treaty highlights growing tensions between political leadership and technical experts regarding AI governance. Khanna expressed skepticism about a recent meeting between the White House, House Speaker Mike Johnson, and American tech executives, calling it a "branding exercise." He argued that companies motivated by profit maximization should not be the sole arbiters of safety rules, comparing the situation to trusting pharmaceutical executives to determine drug safety.
However, the prospect of a binding treaty faces significant political hurdles. President Trump has resisted most regulations, stating on Truth Social that he wants to leave superintelligence "exactly where it is," a position he claimed China shares. House Speaker Mike Johnson has warned against losing the AI race to China, emphasizing a balance between oversight and innovation. While Khanna acknowledges a treaty is unlikely in the near term, he stated his immediate goal is to "shame" Chinese companies and spark a debate among them about safety, noting that US firms are at least engaging in these conversations.
The Bottom Line
Rep. Khanna remains confident that the US will drive AI safety standards, partly due to his belief that many Chinese models are distilled from US models. He advocates for technologists and engineers to lead the debate rather than CEOs or politicians lacking technical expertise. While he ultimately seeks a ban on RSI and monitoring of frontier labs, he proposes an interim working group of experts from both countries to suggest safety recommendations and compliance methods.