The future of open-source AI is at a critical juncture, with the potential for a significant regulatory shift in the next six months. The White House is reportedly considering an executive order to manage open models, which could have far-reaching implications for the industry. This development comes amidst growing concerns about the capabilities of open-source AI models, particularly those originating from China. The author argues that a ban on open-source AI models above a certain capability level would be a mistake, as it would stifle innovation and harm the long-term trajectory of AI development. Instead, the focus should be on addressing the underlying issues of distillation and frontier capabilities, which are driving the push for regulation. The author highlights the role of Anthropic, a company that has been vocal about the risks of Chinese open-source models, in what appears to be a case of regulatory capture. Anthropic's recommendations for policy action are seen as self-serving, as they would benefit from a ban on Chinese models, which would cement their position in the market. The author also discusses the security concerns surrounding model APIs and the potential for distillation to lead to the proliferation of capabilities in the hands of bad actors. The author emphasizes the need for a global agreement on the management of AI model risks to ensure a safe and predictable rollout of open-source AI. In the short term, the author suggests that companies like Microsoft and Meta should release open-weight models to shift the focus away from China's dominance in the open-source space. The author also calls for a coalition of open-source advocates to lobby for the safe rollout of open-weight models and to protect the principles and values of the open-source community.
