California AI Guidance: What State Bar, Courts, & CCPA Require

In September 2025, a California Court of Appeal published an opinion in an otherwise ordinary employment case, Noland v. Land of the Free, LP. The lawyer's briefs contained 23 case quotes. Twenty-one of them were fabricated.

He'd used ChatGPT and other AI tools to help write the briefs, and never checked what came out. The court fined him $10,000 and referred him to the State Bar. It is the largest AI sanction in California so far, and two more instances happened quickly after.

California doesn't have one single AI rule for lawyers, unlike New York’s Part 161 or Florida’s Rule 2.515(d)(2). Instead, there's a stack of stuff: guidance from the State Bar, a first-of-its-kind court rule, a run of sanctioned cases, and a privacy law that could affect how firms use AI day-to-day. Here's what you actually need to know.

The State Bar's Practical Guidance

The State Bar of California put out its first Practical Guidance for the use of generative AI in the practice of law back in 2023. But earlier this year, they updated that guidance to add language about agentic AI.

While this guidance is technically non-binding, it’s still very important for lawyers to understand if they plan to weave legal AI into their workflows. Think of it like a set of best practices rather than a strict rule.

Their Practical Guidance covers the basics:

• Competence. Understand what the tool can and can't do — including the fact that it can just make things up.
• Confidentiality. Don't put client info into a public AI tool without checking it's actually secure.
• Compliance. You must follow all applicable laws, even when using AI tools.
• Supervision. You're responsible for what the AI writes, same as if a paralegal wrote it.
• Communication. Be honest with your client, and let them know how you’re using AI.
• Billing. AI can make you faster. Your client should benefit from that, not just you.
• Candor. Every citation and quote needs to be checked before you file it.
• Discrimination. Understand the inherent bias in many AI tools, and take steps to mitigate depending on the case’s topic.
• Responsibilities. Read up on jurisdiction-specific laws and regulations regarding AI, and make sure to comply with them.

However, these “best practices” may not stay “practices” for long. In 2025, the California Supreme Court directed the State Bar to consider turning this guidance into a rule. And earlier this year, the state bar committee published proposed rule changes for public comment, and that comment period closed in May. 

As of now, the changes still need the California Supreme Court's sign-off. But if they go through, skipping AI verification in California wouldn't just be bad practice — it'd be a disciplinary violation.

Judicial Council Rule 10.430

Separate from anything the State Bar does, California's courts made their own rule for how they use AI. In July 2025, the Judicial Council adopted Rule 10.430 — the first statewide court rule on generative AI in the country.

The gist of it is this: any California court that lets its judges or staff use generative AI had to have a written policy in place by the end of 2025. That policy has to block confidential information from public AI tools, prohibit unlawful discrimination using AI, require someone to catch and fix AI mistakes of all kinds, and flag anything that's entirely AI-written before it goes out. 

Judges have their own version of the same idea, too, since a hallucinated fact in a court order is a much bigger problem than one in a first draft.

2026 CCPA AI Rules

Separate from anything about court filings, California has a privacy law that could affect how firms use AI in their own operations, not just in briefs.

New rules under the California Consumer Privacy Act took effect January 1, 2026, covering something called automated decision-making technology, or ADMT.

If your firm uses AI to make a significant decision about someone — hiring, screening clients, evaluating an employee — these rules apply. Businesses have to tell people when ADMT is being used, let them opt out in most cases, and sometimes give them a way to appeal the decision to an actual human. Firms already using ADMT this way have until January 1, 2027, to fully comply.

None of this touches what you file in court. But it's one more reason to know exactly what your AI tools are doing behind the scenes.

Where Rev Fits In

While the guidelines above don’t regulate evidence itself, every lawyer worth their weight understands that’s what their case actually runs on.

That's where Rev's Investigative Intelligence Platform comes in. Every output ties back to the source, rather than a generated or hallucinated best guess. Here’s what this means in practice:

Litigators can search every deposition in a matter at once and pull the exact page and line before opposing counsel does. Criminal defense and prosecution teams can find the one moment buried in hours of body cam footage before a hearing — cited to the second, not pulled from memory.

California's rules don't require any of this yet. But after three sanctioned lawyers and a state supreme court pushing for tougher rules, the bar for defensible AI use in a California courtroom is only going up. Rev is built for where it's heading.

Learn more about Rev's legal transcription services in California. 

Curious how other states handle this? Check out our post on Texas’ AI law for attorneys.