Vinod Khosla urges US-China treaty to restrict open source AI models

Venture investor Vinod Khosla says open source AI models are more dangerous than closed ones because anyone can run them anywhere without oversight, according to a post by Rohan Paul on X dated November 5, 2026.

Khosla recommends that David Sacks and the White House pursue a treaty with China to constrain open source models from enabling harms such as designing a dangerous virus.

Khosla also proposes limiting open source models to certified clouds, where their use for certain purposes could be restricted. The report is a single social media post summarizing Khosla's views. It does not include a formal proposal, a government response, or any treaty text.