Governor Gavin Newsom Signs Comprehensive AI Safety, Watermarking, and Compute Transparency Act

SACRAMENTO, Calif. — Governor Gavin Newsom signed landmark artificial intelligence regulatory legislation on Monday, enacting comprehensive statutory safety benchmarks, mandatory cryptographic provenance watermarking for synthetic audio and video, and robust statutory whistleblower protections for engineers reporting safety non-compliance at frontier AI labs.

The legislation, negotiated in close consultation with leading AI researchers, open-source developer consortia, and State Senator Scott Wiener, establishes binding safety testing protocols for frontier models trained with over 10^26 integer or floating-point operations (FLOPs) or costing over $100 million to train.

Balancing Tech Innovation with Consumer Protection and Safety With California home to 35 of the world's top 50 artificial intelligence developers, the legislation sets the definitive regulatory benchmark for the global technology industry in the absence of comprehensive federal AI statutes.

"California is the undisputed global epicenter of artificial intelligence innovation, and we have a responsibility to lead the world in responsible AI governance," Governor Newsom said in San Francisco. "This balanced legislation protects consumers from deceptive deepfakes, requires frontier model developers to conduct rigorous catastrophic safety evaluations, and provides complete legal protection for safety whistleblowers, all while safeguarding open-source innovation and startups."

Key Statutory Provisions Under the enacted legislation:

• Mandatory Cryptographic Watermarking: Commercial generative AI platforms must embed invisible, tamper-resistant C2PA provenance metadata into all AI-generated images, audio, and video.
• Frontier Safety Protocols: Developers of mega-scale frontier models must implement written safety and cybersecurity plans detailing red-teaming procedures against catastrophic biological, chemical, or cyber warfare capabilities.
• Frontier AI Whistleblower Shield: Providing statutory protections against non-disclosure agreements for tech employees reporting critical safety hazards to the California Department of Technology and the State Attorney General.

Open-source software advocates commended the legislation for explicitly exempting academic researchers, hobbyists, and fine-tuned open-weight models below training compute thresholds.