California enacts new transparency laws for AI content and chatbots

New transparency requirements

California has officially signed three new bills to regulate how generative artificial intelligence is used and disclosed to the public. These laws are part of a growing movement to ensure that people can distinguish between human-made content and machine-generated outputs. The legislation addresses everything from the images we see online to the voices we hear in advertisements.

One specific law, SB 1000, amends the California AI Transparency Act to cover providers of publicly accessible generative AI systems. This bill removes a previous requirement that systems only be covered if they had over a million monthly users. By removing this threshold, the law casts a wider net over the types of AI technologies that must comply with state transparency standards.

Under these rules, providers must include a 'latent disclosure' in any images, video, or audio that their systems create or alter. A latent disclosure is a piece of information embedded in the content that identifies the provider, the system version, and the date of creation. If a direct disclosure is not technically feasible, the provider must link to a website containing this specific information.

Verifying AI-generated content

To help the public navigate a landscape filled with synthetic media, the state is requiring a new verification system. Covered providers must make a free, publicly accessible 'disclosure verification tool' available to the general public. This tool is designed to let users assess whether a piece of content was created or altered by an artificial intelligence system.

The verification tool must also provide the system provenance of the data used in the content. This means it should show the origin and history of the information that the AI used to generate the final output. The law specifically ensures that the verification tool will not output personal data unless the individual involved has given informed consent.

These requirements are intended to create a standard for accountability in the AI industry. By forcing companies to provide a way to check the authenticity of media, the state hopes to reduce the risk of people being misled by sophisticated digital alterations. This move signals a shift toward making AI transparency a standard feature of digital products.

Rules for synthetic performers

The state is also introducing specific rules for synthetic performers, which are digital figures or voices created through machine learning. These performers are defined as representations that appear to be human but are not based on any specific, identifiable person. The law, known as SB 1050, will specifically target these figures when they appear in advertisements for products or services.

Starting in 2027, any advertisement published in California that prominently features a synthetic performer must include a clear and conspicuous disclosure. This notice must be placed near the performer and remain on the screen long enough for a reasonable consumer to see and understand it. The goal is to prevent consumers from being tricked into thinking a real human is endorsing a product.

This law differs from regulations in other states like New York or Washington. While those states have focused primarily on deepfakes and digital replicas of real people, California's law covers synthetic figures that are not intended to depict a specific individual. It creates a broader category of protection for the general public against deceptive synthetic marketing.

Chatbots and human assistance

Large private businesses with more than $500 million in national gross annual revenue will face new rules regarding customer service chatbots. These companies are prohibited from representing that an AI chatbot is a human being. If a reasonable person could mistake the chatbot for a human, the company must disclose the use of AI in the interaction's medium.

The law also requires these businesses to provide a simple way for customers to request help from a real person. This request feature must be available during at least regular business hours. The company must then make a good-faith effort to connect the customer with a human agent within 15 minutes of that request being made.

If a human agent is not available within that 15-minute window, the business must offer the customer a way to schedule an appointment within one business day. This ensures that while companies can use AI to handle high volumes of inquiries, they cannot completely remove human oversight for frustrated or complex customer issues. These rules aim to balance technological efficiency with consumer rights.

Timeline and compliance

The implementation of these laws follows a staggered timeline to give companies time to adjust their operations. SB 1000, which deals with general AI transparency and disclosure tools, took effect immediately upon its signature. However, the other two major pieces of legislation have a longer lead time for compliance.

Both SB 1050, which regulates synthetic performers in ads, and AB 1609, which governs customer service chatbots, are set to take effect on January 1, 2027. This gives businesses several years to update their internal governance programs and technical infrastructures. Companies will need to audit how they deploy AI and ensure their systems meet the new disclosure standards.

Industry experts suggest that these California laws signal a broader trend in state-level AI and privacy regulation. As more states consider similar rules, companies may find it easier to adopt a single, high standard for AI transparency across the country. For now, businesses are encouraged to review their current privacy and AI governance to prepare for these upcoming legal requirements.