Representative Ro Khanna, ranking member of the House Select Committee on China, has called for a binding, enforceable, and verifiable artificial intelligence treaty between the United States and China. In an opinion piece, Khanna stated that humanity loses if it loses control of AI, arguing that the two nations must move beyond discussion to concrete agreements.
Khanna convened the first congressional hearing last month to address this issue. The event featured experts who recommended that the U.S. and China establish joint technical working groups. These groups would include top scientists from both countries to develop verification technologies that prevent cheating under a future treaty. Khanna described this as an immediate next step following the bilateral dialogue on AI announced by President Donald Trump and General Secretary Xi Jinping.
The representative outlined five pillars for a proposed U.S.-China AI treaty. The first is a worldwide pause on recursive self-improvement, a process where AI develops AI without human intervention. Khanna argued this is the most likely path to humanity losing control of the technology.
The second pillar is a global pause on superintelligence, defined as AI that exceeds human capabilities across all domains. Khanna stated that neither America nor China is ready for such systems.
The third pillar requires kill switches embedded in all advanced models. While noting that scientists disagree on the efficacy of these switches, Khanna said they would provide a chance to stop rogue models.
The fourth pillar mandates that every AI model be tested in an air-gapped sandbox prior to release. This measure aims to prevent catastrophic misuse by criminals, terrorists, or misaligned AI agents.
The fifth pillar focuses on verification. Citing President Ronald Reagan's principle to "trust, but verify," Khanna emphasized the need for inspection regimes to ensure China cannot cheat while its labs race ahead.
Khanna compared the situation to the Cold War, noting that leaders like Reagan and Mikhail Gorbachev recognized that nuclear war would harm both nations. He argued that winning an AI war is similarly catastrophic for civilization.
In addition to international cooperation, Khanna called for domestic action. He urged Congress to pass laws preventing AI companies from prioritizing profit over human safety. He also advised American AI labs to strengthen cybersecurity to prevent China from stealing advanced model weights, which would undermine U.S. leverage in negotiations.
Khanna concluded that the arrival of superintelligence marks a consequential moment for the species, potentially more so than the Industrial Revolution or the development of the atomic bomb. He stated that a global pandemic unleashed by rogue AI would not distinguish between American and Chinese citizens.