An American Answer to China's Open AI Models: A Safe Harbor for Open Weights (manhattan.institute)

In response to China's dominance in open-weight AI models, American AI labs have proposed a legal safe harbor for cooperation, encouraging the release of open AI models while addressing national security concerns. Currently, Chinese models dominate the market, providing developers with easily accessible and customizable tools but posing significant risks related to security, bias, and intelligence obligations under Chinese law. The American approach would involve committing to a predictable schedule for releasing open-weight models that pass independent risk evaluations, while gaining the legal latitude to collaborate on security issues. This strategy is significant for the AI/ML community as it seeks to level the playing field against the heavily subsidized Chinese market, which has grown to account for a substantial share of open model downloads and has been found to produce less secure code. By committing to regular, secure releases of American models, the initiative aims to boost innovation and competition while ensuring the integrity of AI technologies critical for stateside applications. This policy could foster a more dynamic AI ecosystem in the U.S., ultimately extending its leadership in AI across the full technological stack.