California has become the most aggressive state in the country when it comes to AI regulation. More than a dozen new laws either took effect on January 1, 2026, or are scheduled to kick in later this year.

The good news: most of them were designed to regulate AI developers and large companies, not a small business owner using off-the-shelf tools to write emails or answer customer questions.

The confusing news: the language is broad enough that some of these laws may apply to you depending on what you do with AI.

Here's the practical breakdown.

The Law That Matters Most for Small Business Owners Who Build Things

AB 2013 - Generative AI Training Data Transparency Act (effective January 1, 2026)

This law affects you if you develop or substantially modify a generative AI system for public use. If you're a small business that built a custom AI chatbot, trained an AI model on your own data, or created an AI product that others use — this applies to you.

What it requires: public disclosure of what data was used to train your AI system, where it came from, and whether it includes personal information.

What it doesn't require: compliance if you're just using someone else's AI tool (ChatGPT, Claude, Gemini, etc.) and not building anything.

Bottom line for most small business owners: If you're a user of AI tools rather than a builder, AB 2013 doesn't apply to you.

The Chatbot Disclosure Law

SB 1047 / Chatbot rules - If your business uses a chatbot to interact with customers, California law requires you to disclose that customers are talking to a bot, not a human, when they ask.

This is already a common practice, but in California it's now a legal requirement. If your customer service chatbot doesn't identify itself as an AI when directly asked, that's a violation.

Bottom line: Easy fix. Make sure your chatbot identifies itself when asked.

The Deepfake Marketing Rule

California prohibits using AI to create realistic digital replicas of real people in marketing content without their consent. This includes fake celebrity endorsements, AI-generated testimonials using real people's voices or images, and synthetic media featuring identifiable individuals.

Bottom line: Don't use AI to fake customer testimonials or create fake celebrity endorsements. You were probably already not doing this.

The AI Content Labeling Law

SB 942 - California AI Transparency Act (effective August 2, 2026)

Large generative AI providers must label AI-generated content with visible watermarks and metadata. This primarily affects the platforms (OpenAI, Google, Meta) rather than the businesses using them.

However, if you're publishing AI-generated content in California, you should be aware this law exists and that platforms will start embedding labels in AI-created images, audio, and video.

Bottom line: Not your problem directly, but your AI-generated content will start carrying watermarks whether you want it to or not.

The Employment AI Law

If you use AI tools to help make decisions about hiring, firing, promotions, or performance reviews, California has rules about this. You can't use AI systems that produce discriminatory outcomes based on protected characteristics.

Bottom line: If you use AI for HR decisions, make sure the tools you're using have clear anti-discrimination documentation. Ask your AI vendor about this directly.

What California Small Business Owners Actually Need to Do

1. If you use AI tools but don't build them: Review your chatbot setup and make sure it identifies itself as AI when asked. That's probably your only direct compliance action.

2. If you built a custom AI product: Talk to a lawyer. AB 2013 has disclosure requirements that need documentation.

3. If you use AI for hiring: Document which tools you use and ask vendors about their anti-discrimination compliance.

4. Stay tuned: More California AI laws are in the pipeline. The state has made AI regulation a priority, and the rules will keep changing.

California isn't trying to stop small businesses from using AI. The laws are aimed at companies building AI systems. But the broad language means it's worth knowing what you're dealing with.

Sources: Pillsbury Law, Stanford Cyber Policy Center, California Legislature