Microsoft co-founder Bill Gates plans to meet with Chinese President Xi Jinping in November to present proposals aimed at establishing global safeguards for artificial intelligence, including potential restrictions on high-risk AI models.
Gates, who has previously engaged with Xi in 2023—the first foreign business leader received by the Chinese president post-pandemic—outlined his stance in a Wednesday blog post, emphasizing the need for international coordination to address AI's evolving threats. He highlighted recent cybersecurity evaluations where models from OpenAI, Anthropic, and Meta breached real-world websites as evidence of AI's growing capabilities and vulnerabilities.
The proposals include the creation of a new international organization to oversee AI development and deployment, modeled after frameworks used in nuclear inspections, aviation safety, and ozone protection agreements. Gates argued that such an entity could monitor AI's potential to enable biological attacks or synthesize hazardous molecules without stifling innovation across sectors.
In parallel, Demis Hassabis, CEO of Google DeepMind, has advocated for a regulatory body akin to the U.S. Financial Industry Regulatory Authority to conduct national security-focused AI testing. Gates also noted China's existing measures to shield children from AI-driven behavioral addiction, suggesting such policies could serve as a reference for global standards.
Gates' push for AI governance comes amid broader concerns about the technology's risks, which he described as requiring "incredible political management" to navigate safely. He asserted that AI risks surpass those posed by climate change or health crises, framing the issue as the "world's highest priority."
The Microsoft co-founder previously discussed these ideas during the World Economic Forum in Davos in January 2026, where he emphasized the unprecedented scale of AI's current evolutionary impact. His upcoming meeting with Xi Jinping is expected to focus on securing international agreement to restrict the deployment of dangerous AI models, contingent on U.S. leadership in setting precedents.
