US and China Seek AI Safety Talks Amid Chip Export Standoff

U.S. President Donald Trump and Chinese leader Xi Jinping are expected to meet this week to address the growing dangers of uncontrolled artificial intelligence. The discussion comes as both nations compete for technological supremacy while acknowledging that rogue AI systems could pose severe threats to global security and critical infrastructure.

According to reporting from nhregister.com, U.S. Treasury Secretary Scott Bessent revealed that the two countries have proposed a formal mechanism to notify each other of AI incidents that reach national security levels. This proposal aims to create a channel for communication during emergencies, such as major model failures or cyberattacks, even as broader geopolitical tensions remain high.

Shared risks drive diplomatic agenda

Experts note that while competition defines the relationship, both governments share specific concerns about loss of control over AI agents. Senior fellow Sun Chenghao at Tsinghua University argues that dismissing safety debates as purely geopolitical is a mistake, citing China’s own growing attention to malicious use and cyber threats. The focus is shifting toward practical cooperation on how to report and manage these shared risks.

Michael Kratsios, the U.S. science and technology adviser, emphasized the importance of avoiding shared risks as a primary goal for the conversation. He suggested that establishing technical confidence through incident reporting could serve as a stabilizing factor, allowing both sides to manage the potential for AI-enabled attacks on critical systems like finance and telecommunications.

Export controls remain a barrier

Despite the push for safety dialogue, the trade landscape remains contentious. President Trump has repeatedly stated that the United States must maintain a technological edge, while Beijing accuses Washington of pursuing an AI monopoly. A key point of friction is the restriction of access to advanced chips, which Beijing views as an attempt to curb its AI growth.

Analysts indicate that any loosening of these U.S. export controls is unlikely to be part of the upcoming discussions. The divergence in strategic goals means that while the two nations may agree on the need for safety protocols, the fundamental competition for market dominance and technological superiority will continue to shape their interactions.

Building technical trust through dialogue

Vice dean Xiao Qian from the Institute for AI International Governance suggests that discussing emergency communication protocols can help build long-term technical confidence. Samm Sacks of Johns Hopkins University noted that Chinese experts have shown genuine interest in sharing information on cyber incidents, indicating a willingness to engage on specific, manageable topics despite broader political disagreements.

The outcome of these talks will test whether the two superpowers can isolate safety concerns from their wider strategic rivalry. By focusing on concrete mechanisms for incident response, the U.S. and China may find a narrow path toward cooperation that does not require resolving their deeper economic and technological disputes.