The US government signed agreements with Google DeepMind, Microsoft, and xAI to evaluate frontier AI models before public release.
China's 2023 Generative AI rules already require pre-release security assessments and model registration with the Cyberspace Administration of China. China's approach is tied directly to content control and state supervision, while the U.S. approach is framed around national security and cybersecurity. Most importantly, in China, there is a mandatory registration requirement and, in the US, at least for now, this is a voluntary effort.
Will the pre-release review mechanism stay narrow and technical or grow into something closer to a licensing regime? Will it remain voluntary?
[link] [comments]