The United States and China are building a dedicated government-to-government channel on artificial intelligence safety risks, a diplomatic development set to surface during Chinese President Xi Jinping’s visit to Washington. According to reporting from China Tech News, the diplomatic push aims to establish a working group to address shared technical hazards, though both governments are actively managing expectations ahead of the talks.
The emerging dialogue represents a shift in bilateral engagement under President Donald Trump’s second term, moving beyond export-control disputes to focus specifically on systemic technological hazards. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng met on September 16, 2026, to fold AI safety into a broader economic agenda.
What is on the Negotiating Table
While a formal name, charter, or signed text does not yet exist, officials from both nations have floated a specific agenda for the proposed working group. According to China Tech News, the discussions will focus on technical risks rather than commercial dominance. The agenda covers:
- Rogue AI models and agents: Addressing shared concerns about advanced systems acting outside of human-intended bounds.
- Non-state bioweapon risk: Preventing AI applications from lowering technical barriers to biological weapons development by non-state actors.
- Infrastructure and cyberattack risk: Mitigating AI-directed threats targeting critical national infrastructure.
- Incident information-sharing: Establishing a lighter-weight version of threat-intelligence exchanges already discussed between private labs.
- Avoiding system bifurcation: Maintaining open technical communication channels even as both countries decouple major supply chains.
Strict Boundaries on Export Controls
U.S. officials have drawn a hard line around the scope of these discussions. Commerce Secretary Howard Lutnick explicitly ruled out including semiconductor and chip export controls in the AI safety talks. Those commercial and hardware restrictions are being handled through entirely separate diplomatic and regulatory channels.
This deliberate scope decision separates the new safety dialogue from existing trade restrictions, ensuring that negotiations over technical guardrails do not derail broader economic disputes or chip-level oversight measures.
From the 2024 Nuclear Pledge to the 2026 Summit
Direct government talks on AI risk have a limited prior precedent. In November 2024, on the sidelines of the APEC summit in Lima, leaders from both countries affirmed that decisions regarding the use of nuclear weapons must remain under human control rather than automated AI systems. Following that agreement, formal AI-specific talks went dormant.
According to China Tech News, the current restart stems from a Trump-Xi agreement made in Beijing in May 2026. The September 16 meeting between Bessent and He Lifeng served as the first visible follow-through, leading directly into the September 24 Washington summit. Further working-level meetings are tentatively scheduled for November and December 2026 if the summit yields actionable progress.
Despite the high-level engagement, diplomats and analysts remain realistic about the immediate outcomes. U.S. officials indicate that neither government expects a major breakthrough from the September 24 summit itself. A parallel statement reaffirming the 2024 nuclear-control consensus alongside a general statement of shared technological concerns remains the most likely deliverable.