Overview
- Treasury Secretary Scott Bessent said Tuesday that Washington and Beijing have opened a formal dialogue on “superintelligence” to guard against shared AI risks and to avoid misunderstandings between their systems.
- The two sides agreed to exchange best practices for evaluating and testing advanced models before release and to keep human judgment central to oversight of those systems.
- Officials established a direct incident‑alert channel so either country can promptly notify the other about serious AI model failures or dangerous behavior.
- The administration framed the approach as “de‑risk, not to decouple,” pairing risk-management talks with trade steps such as proposed Boards of Trade and Investment to rebalance economic ties.
- A separate, unresolved corporate security issue involving NVIDIA CEO Jensen Huang has prompted a board probe and a shareholder suit and was not addressed in Bessent’s announcement, raising questions about private‑sector ties as diplomacy moves forward.