# Washington and Beijing Initiate AI Stability Dialogue U.S. and China establish communication channels to mitigate AI-related national security risks. - **Source:** [War on the Rocks](https://warontherocks.com) - **Published:** 2026-11-06 - **Topic:** [politics](https://brifing.si/c/politics) - **Original article:** - **On Brifing:** https://brifing.si/s/b345c812-5d43-4ebc-b782-12e40c556f1f ## New Bilateral AI Dialogue U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng met in New York to propose a formal dialogue on AI security. This initiative aims to establish a notification mechanism for national security incidents. The move seeks to transition from opaque competition to greater transparency between the world's leading AI powers. ## Commitment to Human Control Following the meeting between President Trump and General Secretary Xi Jinping, both nations affirmed that AI must remain under human control. This agreement mirrors their 2024 pact regarding nuclear weapons. Both leaders emphasized the necessity of exchanging views on AI risks and benefits to prevent potential misuse and harmful system abuse. ## Super Intelligence Dialogue The United States and China officially agreed to launch the U.S.-China Super Intelligence Dialogue to address risks associated with advanced systems. A bilateral communication channel for incidents will be established, with the next exchange scheduled for November 2026. This framework focuses on managing uncertainty rather than imposing restrictive arms control measures. ## Strategic AI Competition Both nations view AI as a central pillar of their strategic rivalry, with leaders acknowledging that dominance in this field is critical. China's current five-year plan prioritizes seizing technological heights. AI provides transformative military advantages, including improved intelligence collection, autonomous maneuver capabilities, and enhanced command and control functions for modern battlefield operations. ## Nuclear Escalation Risks Integrating AI into nuclear command and control systems creates significant instability. Research indicates that large language models, including China's DeepSeek, may bias decision-making toward escalatory options during simulated crises. Furthermore, reliance on AI for early warning systems risks distorted intelligence analysis, which could inadvertently trigger a conflict between the two nuclear-armed states. ## Limitations of Arms Control Formal arms control for AI remains unlikely due to the expiration of the New START treaty and the emergence of a tripolar nuclear age. As a dual-use, general-purpose technology, AI presents unique challenges for verification and disclosure. Great powers currently show little appetite for constraints that might limit their future military advantages. ## Cold War Precedents Confidence-building measures, such as the 1963 hotline and the 1972 Incidents at Sea Agreement, offer historical templates for managing rivalry. These voluntary, non-binding measures helped the U.S. and Soviet Union avoid accidental escalation. Modern efforts aim to replicate this transparency to mitigate the risks of inadvertent conflict during periods of high tension. ## Historical Successes and Failures The 1975 Helsinki Final Act and the 2011 Vienna Document demonstrate how communication channels can evolve over decades. However, these measures are fragile; Russia’s 2023 withdrawal from the Vienna Document highlights their voluntary nature. While they do not deter aggression, they provide early warning signs when parties deviate from established behavioral expectations. ## Absence of Crisis Catalysts Unlike the Cuban Missile Crisis, which forced the U.S. and Soviet Union to institutionalize communication, no comparable AI-triggered crisis has occurred. Without a shared, vivid sense of immediate danger, it remains unclear if both sides fully agree on the specific nuclear risks posed by AI beyond the general principle of human control. ## Conceptual Definition Gaps The U.S. and China lack consensus on key AI definitions, including what constitutes a lethal autonomous weapon. China sets a high threshold for intervention, while U.S. policy remains more permissive regarding autonomous systems. These divergent interpretations complicate efforts to negotiate specific restrictions or shared standards for military AI applications and deployment. ## General-Purpose Technology Challenges AI is not a discrete weapon system but a general-purpose technology with vast civilian and military applications. This complicates efforts to isolate and regulate specific AI-nuclear intersections. Because many AI capabilities are essential for broader military competition, both nations are reluctant to accept constraints that might weaken their overall strategic and technological posture. ## Private Sector Influence Private firms control the resources and innovation necessary for AI development, making them integral to any confidence-building effort. Their interests often diverge from government objectives, particularly regarding intrusive inspections or sensitive information sharing. Future stability depends on whether governments can align these corporate equities with national security requirements for transparency. ## Sources covering this story - https://warontherocks.com/can-washington-and-china-build-confidence-around-ai-and-stability/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._