Gates argued that risk management in artificial intelligence requires cooperation between the two countries. Since both lead AI development, setting divergent standards for handling risk would blunt the effect of either.

The areas he cited are reported to include methods of safety evaluation, sharing information on serious failures, and preventing misuse in biological and chemical domains — all fields where common interest exists without colliding directly with competitive advantage.

Why it is hard

The conditions are unfavorable. Semiconductor export controls, restricted access to model weights and limits on researcher mobility have narrowed the channels of technical exchange. Safety research ultimately requires access to models, and that access is itself controlled.

Nuclear safety and infectious-disease information sharing are the usual precedents. In both, competition coexisted with cooperation, and keeping a minimal channel open is credited with limiting the spread of accidents.

What is actually moving

International bodies and conferences on AI safety are multiplying, though most stop at declarations of principle rather than binding agreements. Agreement on testing methods through standards organizations is seen as the channel where practical progress happens.

Sinophone Bridge Times does not relay the remark as one camp's position. What is confirmed is that a figure at the technological frontier publicly stated the need for cooperation; what institutions might follow remains open.