News -

New AI Transparency Law in California, USA

Partner, Rotem Perelman-Farhi, and Adv. Dr. Laura Jelinek in a piece on the new AI Transparency Law in CA – from a legal perspective.

On September 28, 2024, the California governor signed into law bill AB 2013, the so-called “AI transparency act”. This new law requires generative AI developers to post certain documentation on their website regarding the data used to train their AI systems. The law applies to generative AI released on or after January 1, 2022, and developers must comply with its provisions by January 1, 2026.

Who is subject to the law? The law applies to AI “developers”, i.e. any person or entity that either develops a generative AI technology or service or that “substantially modifies it,” and that make the AI system available for use in California. The scope is therefore rather broad and means that service providers that engage in material retraining or fine-tuning of existing generative AI models also need to comply with the law.

Scope of the Law: The law regulates “generative artificial intelligence,” which is defined as AI “that can generate derived synthetic content, such as text, images, video, and audio, that emulates the structure and characteristics of the artificial intelligence’s training data.”

What do developers need to do in order to comply? If a developer makes a generative AI system publicly available to Californians, it has to disclose certain documentation regarding the data used to train the system or service, among others:

  • The sources or owners of the datasets, and whether the developer purchased or licensed the datasets;
  • an explanation how the data is used in the AI system or service;
  • whether the datasets include any copyrighted or trademark-protected data;
  • the time period during which the data was collected; and
  • whether the datasets include “personal information” or “aggregate consumer information” as those terms are defined under the California Consumer Privacy Act.

Exemptions: AB 2013 does not apply to generative AI systems or services that are (1) solely designed to ensure security and integrity, (2) solely intended for the operation of aircraft within the national airspace, or (3) developed for national security, military, or defense purposes, and are made available only to a federal entity.

Other AI laws in California

Together with AB 2013, several other AI-related bills have been signed into law in California. Several of these focus on the use of AI in healthcare and in education; another bill requires watermarking of AI-generated content under certain circumstances.

The governor vetoed, however, SB 1047, known as the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act. This bill, supported by Elon Musk and Anthropic, and opposed by OpenAI, Meta and Google, had proposed the introduction of new safety requirements for large-scale AI models. Developers of large AI models, among others, would need to establish full shutdown capabilities, outline safety and security protocols to prevent significant harm to public infrastructure, and adhere to specific audit requirements.

Download as PDF
Share:

Media