How to Make the U.S.-China AI Race Less Dangerous
The United States and China aren’t ready to slow down but they still have plenty to talk about.
The recent meeting between Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng presents a challenging moment for the United States and China to discuss artificial intelligence. President Donald Trump has dismissed public warnings about AI as a hoax, and China's state-owned newspaper accused U.S. AI companies of hindering China's progress.
Trust is lacking between the two nations, and AI systems are advancing rapidly, regardless of diplomatic efforts. However, the potential dangers of these systems cannot be ignored. A dialogue on AI safety is necessary, but it must avoid the pitfalls of setting goals too low or too high. Instead, a balanced goal would be for the U.S. and China to pursue AI safety in parallel and demonstrate progress.
The talks should focus on technical exchanges that make each country's AI systems safer on their own terms. They should also establish a protocol for communicating about emerging AI incidents before they escalate into international crises. These steps would represent significant progress and are more valuable than lofty statements or unverifiable commitments.
Both countries should study potential AI risks, such as systems that could help amateurs develop bioweapons or evade human oversight. American AI companies have been researching these risks for years and have begun to appear in Chinese technical standards. Exchanging best practices and new insights on testing and mitigating these risks can make the systems built in both countries safer.
This diplomacy is more about establishing shared safety practices and procedures for a technology that neither government fully understands or controls. It is not about one side trusting the other but recognizing that neither benefits if the other loses control of a powerful system. Serious, technically-informed diplomacy is required to address potential incidents, such as AI models escaping control or attacking U.S. or Chinese entities.
The talks should begin without preconditions or concessions on broader technology competition issues. If Beijing seeks to make AI safety progress conditional on relief from U.S. export controls on advanced chips, Washington should keep these issues separate. The goal of AI safety diplomacy is not to negotiate an arms-control treaty but to develop shared safety practices and procedures for a rapidly advancing technology.
Written by urgent.news from Time's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
