US and China to Hold First AI Safety Talks of Trump Second Term

The United States and China are preparing for their first dedicated bilateral AI safety talks since the start of Donald Trump's second term, scheduled for mid-September 2026. The dialogue aims to establish critical guardrails for the rapid advancement of artificial intelligence before a high-level diplomatic visit later in the month.

US Treasury Secretary Scott Bessent will lead the American delegation for the mid-September meetings. The talks are designed to address the systemic risks posed by frontier AI, with a specific focus on preventing AI-driven cyberattacks and the proliferation of powerful models to non-state actors. This diplomatic effort precedes a scheduled visit by Chinese President Xi Jinping to the United States on September 24, 2026.

The Path to Dialogue

These upcoming talks follow a May 2026 summit in Beijing between President Donald Trump and President Xi Jinping, where AI governance was a confirmed topic of discussion. The engagement builds upon a broader history of AI safety efforts, including the first official talks held in Geneva in May 2024 and the subsequent 'Singapore Consensus' on AI safety. While the two nations have historically diverged—with the US emphasizing national security and chip restrictions and China favoring technical discussions over political debate—the current trajectory suggests a renewed push for formal coordination.

Global Security Implications

As the world's two leading AI superpowers, the ability of the US and China to agree on safety standards is viewed as critical for global governance. The stakes involve more than just technical benchmarks; failure to establish "redlines" regarding frontier AI could significantly increase the risk of AI-enabled cyberwarfare. Furthermore, the uncontrolled proliferation of dangerous capabilities remains a primary concern for Western security officials.

Adding to the tension, former OpenAI executive Dean Ball has warned that China's approach to open-weight AI could lead to a state of "full AI communism," powered by government-controlled data centers. This perspective highlights the ideological divide that complicates the technical goal of safety cooperation.

Next Steps

Observers will be watching to see if the mid-September talks produce concrete agreements on risk mitigation or merely serve as a diplomatic prelude to President Xi's visit on September 24. Key points of contention are expected to include the definition of "frontier models" and the balance between open-source development and national security restrictions. Whether these talks can move beyond general governance discussions to create enforceable safety protocols remains to be seen.