High-Level Overview

What are the core concepts of California SB-942?

California SB-942 aims to increase transparency about the detection and provenance of AI-generated content in the state of California.

The bill requires providers of AI that generates image, video, or audio content to:

1) create and publish an AI detection tool that can be queried to assess whether the organization’s generative AI was used to create the content in question

2) to watermark AI-generated content

Scope

Who does California SB-942 apply to?

The California AI Transparency Act applies to any business that provides a generative AI system within California that creates image, video, or audio content (including content that combines these media types).

Businesses with less than one million monthly users or visitors on average over the preceding 12 months are not in scope. For purposes of this bill, in-scope businesses are referred to as “covered providers.”

Compliance Requirements

What are the compliance requirements of California SB-942?

The Act imposes several compliance requirements:

1) AI content detection tool: Covered providers must make available an AI detection tool that a person can use to query to what extent a piece of text, image, audio, video, or multimedia content the covered provider’s AI was used to create (in whole or in part) the content.

The AI detection tool must be publicly accessible via a URL on the provider’s website and mobile app, as applicable.

The tool must allow the user to upload content directly or submit a URL.

The tool must have an application programming interface (API) that allows the user to query the tool without accessing the provider’s website.

The tool must allow a person the ability to submit feedback if they believe the tool is not correctly identifying content created by the provider, and the provider should incorporate this feedback to improve the tool.

The provider is not allowed to disclose personal information about who used generative AI to create content, nor are they allowed to collect such information.

2) Watermarking: In AI-generated content, the provider must include a visual disclosure that identifies the content as AI-generated, is difficult to remove, and includes the identity of the generative AI used and the date and time the content was created.

The provider must also include an imperceptible disclosure marking the content as AI-generated that is permanent or difficult to remove.

Lastly, in contracting terms with downstream users of the generative AI, the covered provider must include clauses that prohibit the removal of a required watermark and terminate access when it’s believed that a downstream third party did remove a watermark.

Non-Compliance Penalties

What are the non-compliance penalties for California SB-942?

Offending companies face civil penalties of $5,000 per violation.

For violations of the bill’s disclosure requirements by third-party licensees, state authorities may bring civil action for injunctive relief and attorney costs.

Status

When does the California AI Transparency Act go into effect?

The bill passed the California legislature in August 2024 and it is headed to the desk of Governor Newsom.

Update: Governor Newsom signed the Act into law September 20th, 2024. The CA AI Transparency Act will enter effect January 1st, 2026.

Steps To Compliance

How can organizations ensure compliance with the California SB-942?