The short version
- Representative Ro Khanna has formally requested an assessment of AI safety protocols from US intelligence agencies and major Chinese technology firms.
- The proposal aims to establish a bilateral treaty banning recursive self-improvement in artificial systems, though political support remains divided.
- Critics within the administration and Republican leadership argue that strict regulations could hinder innovation and disadvantage American competitiveness.
A new legislative push for international cooperation on artificial intelligence safety has emerged from Capitol Hill, led by Representative Ro Khanna of California. As President Donald Trump prepares to host technology executives in Washington, the top Democrat on the House Select Committee on China is advocating for a formal treaty between the United States and China. The goal is to prevent catastrophic outcomes associated with advanced AI systems, despite significant geopolitical tensions between the two nations.
Khanna recently sent letters to the Office of the Director of National Intelligence and several prominent Chinese artificial intelligence companies, including DeepSeek, Alibaba, and Moonshot AI. These communications seek specific information regarding how effectively the US could respond if an AI laboratory lost control of its autonomous agents. The representative also asked for documentation on efforts to achieve superintelligence and recursive self-improvement, a milestone where systems can enhance their own capabilities without human intervention.
The inquiry specifically probes whether these companies have implemented safeguards or kill switches for their technology. Khanna further inquired if the firms would agree to inspections by non-governmental agencies should a treaty be established. The letters also request an evaluation of the Chinese government’s approach to assessing catastrophic risks posed by artificial intelligence. As of now, neither the intelligence agency nor the Chinese companies have responded to requests for comment regarding these inquiries.
This diplomatic effort follows a recent meeting between President Trump and Chinese President Xi Jinping at the White House. While both leaders agreed to maintain dialogue on AI, Khanna characterized their engagement as insufficient given the urgency of the risks involved. He argues that current discussions fail to adequately address the potential for humanity to lose control over its technological destiny. The representative believes that while American and Chinese interests diverge on many issues, preventing AI-related disasters is a shared necessity.
Khanna has expressed skepticism regarding a planned meeting between the White House, House Speaker Mike Johnson, and American tech executives. He described the gathering as a branding exercise designed to reassure the public rather than a substantive policy discussion. Drawing a parallel to pharmaceutical regulation, he questioned whether profit-driven companies should be trusted to define safety standards for technologies that could pose existential threats.
The proposed treaty would ban recursive self-improvement and establish monitoring mechanisms for frontier AI labs. However, achieving such an agreement faces substantial political hurdles. Republican lawmakers, including House China Committee Chair John Moolenaar, have voiced doubts about the reliability of Chinese commitments to any such pact. Meanwhile, President Trump has indicated a preference for leaving the current state of AI development unchanged, suggesting that regulatory intervention is not a priority for his administration.
Speaker Johnson emphasized the need to balance oversight with innovation, warning against measures that might cause the United States to fall behind in the global AI race. This perspective aligns with broader Republican concerns about maintaining technological competitiveness. In contrast, some leading US AI companies have called for coordination with China to slow the pace of development, highlighting a divergence between industry leaders and certain political factions.
Khanna acknowledges that Chinese firms are unlikely to provide full transparency or compliance. He noted their lack of response to previous requests for commitments to slow down development alongside American counterparts like OpenAI and Anthropic. His strategy involves using these public inquiries to spark debate among Chinese technologists about safety practices. He remains confident that the US will eventually drive global safety standards, partly because many Chinese models are derived from American technology.
The representative advocates for a working group comprising experts from both countries to develop safety recommendations and compliance methods. He argues that current political leaders are out of touch with the technical realities faced by scientists and engineers. By shifting the debate away from profit-motivated CEOs and toward technical experts, Khanna hopes to create a more informed policy environment. The outcome of these efforts will likely depend on whether bipartisan support can be built around the need for international AI governance.
As the dialogue continues, the tension between national security concerns and technological innovation remains central to the debate. While some view regulation as a necessary safeguard against uncontrollable AI systems, others see it as an impediment to economic growth and strategic advantage. The coming months will reveal whether these competing priorities can be reconciled through diplomatic channels or if they will deepen existing divisions in US-China relations.
Sources behind this briefing
Go to the original reporting
- The Verge↗Will Chinese AI companies slow down? A top House Democrat wants answers
