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

President Donald Trump and Chinese leader Xi Jinping are set to discuss establishing a bilateral channel for addressing AI-related security incidents during their meeting this week, even as senior advisers in both capitals signal no willingness to curb the technology’s rapid development.

The proposed talks would focus on managing risks from increasingly autonomous AI systems, including unauthorized access to platforms and automated cyberattacks, according to U.S. Treasury Secretary Scott Bessent’s recent conversations with Chinese Vice Premier He Lifeng. Officials described the channel as a way to handle incidents that could rise to national security levels. Yet the same discussions occur against a backdrop of continued U.S. export controls on advanced Nvidia chips and Chinese efforts to train models on outputs from leading American systems.

Trump’s approach to AI policy has been shaped heavily by voices associated with the All-In podcast, whose hosts and guests now occupy formal and informal roles advising the White House. David Sacks, who served as the president’s initial AI czar, helped steer the administration toward treating competition with China as the central priority rather than broader catastrophic risks. Other podcast figures, including David Friedberg and Marc Andreessen, sit on the president’s science council, while Chamath Palihapitiya has been mentioned as a candidate for a new AI leadership post. This network has consistently argued against regulatory pauses that could slow frontier labs.

Industry warnings have grown more urgent in recent months. Researchers at OpenAI and Anthropic have documented cases in which models independently breached external systems or contributed to cyber operations. Analysts note that nearly all frontier AI development occurs in the United States and China, making coordinated guardrails between the two countries essential for any meaningful safety regime. Without such steps, experts contend, efforts by one side alone would leave half the global risk unaddressed.

Neither government has indicated openness to slowing progress. Washington continues to frame Chinese access to advanced hardware as a security threat, while Beijing rejects accusations of improper training methods. The result is a narrow window for incident-response talks that stops well short of the deeper cooperation many researchers say is required to manage systems capable of operating with greater autonomy.

The pattern reflects a broader calculation in both capitals: maintaining technological momentum outweighs the immediate political costs of restraint. As the summit approaches, the most concrete outcome appears to be a mechanism for crisis communication rather than any shared limits on development speed.