“The proposed AI standards body would assess models on risks to cybersecurity, national security, and biological threats, with benchmarks updated quarterly.”