Bill Gates is pushing for a new international body to handle the most dangerous uses of artificial intelligence. In his view, individual governments and AI companies cannot manage highly capable models on their own. He has suggested a framework that would monitor high-risk AI capabilities instead of slowing down normal development.
Safety as a global issue
Gates has compared his idea to nuclear inspection and aviation safety. Both are systems where countries accept outside checks because a failure in one place can affect everyone. He thinks AI should be handled the same way. The focus would be on skills like designing dangerous biological molecules or finding software vulnerabilities at scale.
That concern has become more concrete recently. OpenAI said this month that assessments of a future model showed major progress in cybersecurity and agentic coding. Under its own Preparedness Framework, the company could no longer rule out critical cyber capabilities. That kind of statement helps explain why Gates believes the window for voluntary action may be closing.
The China factor
Gates plans to travel to China in November and hopes to meet President Xi Jinping. He told Reuters that China might agree to limits on releasing particularly hazardous AI models if the United States takes the lead on safety. It is a big condition, considering the current relationship between Washington and Beijing. But Gates thinks U.S.-China coordination is essential. If one country bans risky AI while another allows it, development could simply move to the weaker regulatory area.
Jobs and economic risk
His worries are not limited to security. Gates has also warned that AI could replace white-collar and blue-collar jobs faster than earlier technologies did. He has floated ideas such as stronger safety nets, taxes on AI-driven labor replacement, and reserving certain jobs for human workers. Some of those proposals are controversial. They also show that Gates sees AI as more than a technical problem.
None of this will be easy. AI is now a strategic issue, and cooperation between geopolitical rivals is rare. Still, Gates seems willing to push the conversation forward. Whether that leads to actual global AI control remains to be seen.
