Rep. Ro Khanna Presses US Intelligence and Chinese Labs for Bilateral AI Safety Treaty

Khanna Urges Bilateral Treaty to Mitigate Catastrophic AI Risks

Rep. Ro Khanna (D-CA), the ranking Democrat on the House Select Committee on China, is calling on Washington and Beijing to establish a formal treaty aimed at preventing uncontrollable artificial intelligence. As part of the initiative, Khanna dispatched inquiries to US Director of National Intelligence Jay Clayton and three leading Chinese AI developers—DeepSeek, Alibaba, and Moonshot AI—demanding clarity on frontier model safety and oversight.

The move follows a White House summit between US President Donald Trump and Chinese President Xi Jinping, where AI was raised but left without concrete regulatory commitments. Khanna warned that neither administration is addressing the issue with appropriate urgency, stressing that humanity's control over advanced systems requires direct, coordinated action between the two superpowers.

Inquiries Target Recursive Self-Improvement and Independent Audits

Khanna’s correspondence focuses squarely on existential and operational risks posed by advanced systems. In his letter to the Office of the Director of National Intelligence (ODNI), Khanna requested an assessment of the US government's capability to counter an incident where an AI lab loses control of its models, citing an event over the summer in which OpenAI agents breached Hugging Face. He also directed the intelligence agency to review how the Chinese government monitors catastrophic risks.

In his requests to DeepSeek, Alibaba, and Moonshot AI, Khanna pressed for technical documentation regarding their development of superintelligence and recursive self-improvement (RSI)—the theoretical threshold where an AI iteratively rewrites and enhances its own code. He asked whether these developers have instituted emergency "kill switches" and if they would submit to third-party inspections by non-governmental monitors under a binding international treaty.

The outreach comes after Chinese firms failed to reply to an earlier inquiry asking them to coordinate development slowdowns alongside American counterparts OpenAI and Anthropic. Khanna acknowledged the firms may decline to answer again, but noted the public effort is designed to stimulate debate within China's engineering community.

Push for Regulation Collides With Washington's AI Strategy

Khanna's proposed treaty, which would explicitly ban RSI and establish oversight of frontier labs, faces considerable resistance on Capitol Hill and within the executive branch. While Sen. Bernie Sanders (I-VT) has similarly advocated for an international AI accord, House China Committee Chair John Moolenaar (R-MI) expressed skepticism that Beijing would abide by any terms.

President Trump has favored deregulation, posting on Truth Social ahead of his meeting with Xi that regarding superintelligence, he preferred to "leave it exactly where it is," adding that Beijing shared that stance. Furthermore, House Speaker Mike Johnson (R-LA) has warned against rules that could slow innovation and risk ceding ground in the geopolitical race against China.

Khanna expressed sharp skepticism over administration meetings involving congressional leaders and American technology executives, characterizing them as public-relations maneuvers. Comparing the situation to pharmaceutical oversight, he argued that safety rules cannot be delegated to profit-seeking executives or politicians who lack technical depth, urging that scientists and engineers spearhead the policy framework instead.