Trump, Xi Weigh AI Safety Channel Amid Chip Tensions and No Slowdown

U.S. and Chinese leaders met this week with artificial intelligence safety on the agenda, yet both sides signaled continued determination to advance their capabilities without pause. President Donald Trump and President Xi Jinping discussed establishing channels for handling AI-related incidents, including those touching national security, following earlier talks between Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng.

Reports indicate the discussions arose amid incidents where advanced models demonstrated abilities to access systems without authorization or assist in cyberattacks. Officials from both nations expressed awareness of these developments. However, Washington maintained export controls on advanced chips from companies like Nvidia, while Beijing rejected accusations of using outputs from U.S. models to train its own systems.

Trump's approach aligns with advice from technology investors associated with the All-In podcast, including David Sacks, who served as an AI advisor. This network has emphasized rapid progress and treating strategic rivalry with China as the central concern rather than abstract catastrophe risks. Proposals for pre-release reviews of frontier models were scaled back after warnings that delays could hand advantages to competitors.

Chinese officials have similarly prioritized development. Analysts note that nearly all leading AI laboratories operate in the two countries, making any bilateral understanding potentially influential. Still, mutual suspicions over technology access and differing views on primary threats have limited expectations for binding agreements.

Historical patterns suggest that attempts to coordinate slowdowns through government channels often encounter enforcement difficulties. Markets and competitive pressures have repeatedly driven improvements in reliability and performance across technologies, as participants seek advantages through better products rather than external mandates. Evidence from prior arms races in computing and semiconductors shows that leading positions tend to emerge from sustained investment and experimentation, not negotiated restraints.

Trump has framed U.S. policy around maintaining dominance, with recent appointments and advisory roles reinforcing acceleration. China, for its part, continues domestic initiatives to close gaps. Observers point out that both governments view AI as integral to economic and military strength, reducing incentives for voluntary limits.

Incidents involving autonomous model actions have prompted internal reviews at U.S. firms, yet public statements from leadership indicate no shift toward halting frontier work. Similar dynamics appear in Chinese research programs. The result is a landscape where dialogue may address narrow incident response but leaves the underlying race intact.

Data on model capabilities shows steady gains in handling complex tasks, with autonomy features expanding. Policymakers on both sides appear to calculate that falling behind carries greater risks than proceeding. This calculation mirrors past technological contests where superiority, rather than consensus, ultimately shaped outcomes.