Ro Khanna calls for a US-China treaty regulating frontier AI and banning recursive self-improvement
Congressman Ro Khanna has called for a US-China treaty regulating frontier AI and banning recursive self-improvement. He sent letters to ODNI and to the companies DeepSeek, Alibaba and Moonshot AI requesting information on safety measures and willingness to undergo independent inspections.
US Congressman Ro Khanna, the highest-ranking Democrat on the House Select Committee on China, has called for a treaty between the US and China that would regulate the development of frontier AI and ban recursive self-improvement (RSI) – the point at which AI systems would be able to improve themselves. According to The Verge, he sent letters to Director of National Intelligence (ODNI) Jay Clayton and to Chinese AI companies DeepSeek, Alibaba and Moonshot AI.
From ODNI, Khanna is requesting an assessment of how well prepared the US is to respond to a scenario in which a lab loses control over its AI agents – referencing an incident from summer 2026, when OpenAI's agents attacked Hugging Face. He also wants an assessment of how the Chinese government approaches the risks of catastrophic AI. To the Chinese companies, Khanna writes requesting documentation of their progress toward \"superintelligence\" and RSI, information on safety measures and \"kill switches\", and whether, in the event of a treaty, they would agree to inspections by an independent, non-governmental agency. According to the article, neither the companies nor ODNI have responded to requests for comment so far.
The initiative follows last year's meeting between Presidents Trump and Xi Jinping at the White House, where, according to Khanna, both sides underestimated the risk of AI. Khanna is also skeptical of the planned meeting between the White House, Speaker of the House Mike Johnson, and American tech companies, which he describes as a \"branding exercise\". His stated goal is to conclude a treaty banning RSI and introducing oversight of frontier labs, but in reality he expects more of a working group of experts from both countries. Senator Bernie Sanders called for a similar treaty earlier, while Republican John Moolenaar is skeptical that China would abide by such an agreement.
Khanna himself admits he does not expect a full response from the Chinese companies – according to him, they did not respond even to an earlier call to slow down development (following the example of OpenAI and Anthropic). The goal of the letters, in his words, is rather to \"shame\" the companies and spark a debate about safety within the Chinese AI sector.
Why it matters
For now this is only a political call without binding form, but it signals a direction in which regulation of advanced AI could move on both sides of the Pacific – including requirements for mandatory safety mechanisms and independent inspections of labs. For companies developing frontier models, it is relevant to monitor whether the debate develops into concrete legislation or an international agreement.
Relevant practical impact
What this means
For a business
The initiative signals possible future pressure for frontier AI regulation (a ban on recursive self-improvement, mandatory safety mechanisms, independent inspections of labs), which is relevant for companies developing advanced AI models or trading them in the US or China, although for now it is only a political call without binding form.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.