San Francisco, United States: On Fri, Sep 29, 2023, California enacted the Transparency in Frontier Artificial Intelligence Act, marking it as the first state in the United States to legislate on regulating advanced AI technologies.

AI Regulation Overview

This legislation mandates that developers of the largest frontier AI models, which are highly advanced systems surpassing existing benchmarks, must publicly report their adherence to national and international frameworks and best practices during development. This includes reporting significant incidents such as large-scale cyber-attacks, fatalities, substantial financial losses, and other safety-related events caused by AI models. The law also incorporates whistleblower protections.

Impact on AI Governance

Located in a hub for major AI companies, California’s legislation could influence global AI governance and user implications worldwide. An earlier draft of the bill proposed more stringent measures like kill switches and third-party evaluations, which faced opposition due to concerns about stifling innovation. The final law reflects a compromise, focusing on disclosures without enforceability.

High-Risk Models Not Covered

The law covers only the largest models, excluding smaller but potentially high-risk models. This omission contrasts with the European Union’s AI Act, which includes obligations for both large and high-risk models. Analysts emphasize the importance of accountability, though the current California law does not hold developers liable for crimes committed by AI models but only requires the disclosure of governance measures applied.

Balancing Regulation and Innovation

California’s law aims to protect users while fostering innovation. By focusing on the largest models, it avoids imposing burdens on startups. The legislation also establishes a public cloud computing cluster to support AI infrastructure for startups. The approach serves as a starting point for potential future regulation as more research and testing on high-risk models are conducted.

Waivers to Regulations

Other states, including Colorado, are also pursuing AI legislation, while federal lawmakers remain cautious about national regulation. A proposed bill by Senator Ted Cruz suggests allowing AI companies to seek waivers from regulations perceived to hinder growth, highlighting the ongoing debate between regulation and innovation.

In conclusion, while California’s law is a preliminary step towards AI regulation, it may set the stage for more comprehensive oversight as the field evolves.