SB 574, Translated for Every Lawyer: What California’s New AI Law Means for Your Law Firm

Estimated reading time: 6 minutes

On August 31, 2026, the California Legislature finished its work on Senate Bill 574, the first law in the country written specifically to govern how attorneys and arbitrators use generative AI. The bill was presented to Governor Gavin Newsom on September 9. He has until September 30 to sign it or veto it.

If he signs it, California becomes the first state with a real rulebook for AI in the practice of law. Most law firms have never read a bill like this. Most managing partners do not have time to. So let’s translate it.

Here is SB 574, one idea at a time, translated out of legislative language and into something every lawyer, from first year associate to managing partner, can actually use.

You may also be interested in our article on The Reimagined Law Firm: https://www.iffelinternational.com/law-firm-ai-operations-growth-governance/

The lawyer still has to do the work

Imagine a kid brings a robot to help with homework. The robot can look things up, organize notes, even suggest an answer. But when the teacher asks “who solved this problem,” the answer has to be the kid, not the robot.

That is the first and most important rule in SB 574. A lawyer can use AI to help with research, drafting, or organizing a case. But the lawyer cannot hand the robot the pencil and walk away. The judgment, the responsibility, and the final decision stay with the human.

Secrets stay secret

If you tell your robot helper a family secret, and the robot tells the secret to a stranger, that is a problem. Client information (financial details, medical history, family disputes, anything private) cannot be typed into a public AI tool where anyone outside the firm might see it.

SB 574 says confidential information can only go into an AI system if access is locked down to the attorney and the specific people who are allowed to see it. Convenience does not get to trump confidentiality.

If the robot makes something up, the lawyer has to catch it

Sometimes AI tools invent things that sound true but are not. Lawyers call these “hallucinations.” A fake court case. A quote nobody ever said. A law that does not exist.

SB 574 says the lawyer has to check the AI’s work and fix anything that is wrong before it goes anywhere near a client or a courtroom. Trusting the robot’s homework without checking it is exactly what the bill is trying to stop.

Every citation gets a human check, not just the AI ones

Here is a detail people miss. The bill does not only require lawyers to verify citations that came from AI. It requires personal verification of every citation in a court filing, period. If an attorney’s name is on the document, that attorney has personally read and confirmed every case cited inside it, no matter where the citation originally came from.

Courts get told when AI was used

If a document is filed in court, the attorney has to disclose that generative AI played a role in preparing it. The bill does not spell out exactly what that disclosure has to look like, and reasonable people are still debating how it should work when AI only touched a small part of a document versus a full first draft. That detail will likely get sorted out after the bill becomes law.

The public gets a “please consider telling them” rule

For content shared publicly (not filed in court, just published), the bill asks attorneys to consider whether they should disclose AI’s role. It is a softer rule than the court disclosure requirement, but it points in the same direction: know what your AI touched, and be ready to say so.

Arbitrators cannot let AI make the call

The bill also covers arbitrators, the people hired to settle disputes outside of court. An arbitrator cannot hand a decision over to AI, and cannot rely on AI-pulled information that sits outside the official record without telling the people involved in the dispute first.

What happens next

SB 574 does not carry its own penalties. If it becomes law, it gets enforced the way most attorney conduct already gets enforced: through court sanctions and the State Bar’s disciplinary process. Which brings us to the part every managing partner actually needs to hear.

Law firms are already at risk, and it is a challenge we are tracking closely

This is not a hypothetical. California’s State Bar has already brought discipline cases this year over AI-related citation problems, including a suspension tied to a filing built on cases that did not exist. Courts around the country have sanctioned attorneys for the same reason, and the number of these incidents keeps climbing as more firms adopt AI faster than they build the guardrails to go with it.

That is the gap Iffel International was built to close. We are not watching this trend from the sidelines. We have investigated the how in automating law firm operations with the discipline SB 574 is asking for built in from day one: human oversight on every AI-assisted output, documented review before anything reaches a client or a court, and a clear record of what AI touched and what a person verified.

Part of how we do that is by building on open trust infrastructure like the TIP protocol (Trust Identity Protocol), which lets content carry a signed, verifiable label of its origin: written by a human, assisted by AI, or generated by AI. It is exactly the kind of mechanism SB 574’s disclosure requirement is reaching for, proof instead of a promise.

We believe and you can be the judge that Iffel International is bringing these pieces together, verified AI output, documented human review, and signed content provenance, into one integrated system for law firms ahead of the market. Not outsourced oversight. Not a comforting title with no mechanism behind it. Trust and integrity, engineered in. We study your operations footprint first before deploying any AI automation.

Under California’s standard rule, most new laws without an urgency clause take effect on January 1 of the following year, which would put SB 574 on track to apply starting January 1, 2027, if signed. Either way, the clock is already running. We would rather help you have real answers to these questions now, not after the first complaint lands.

---

Hema Dey is the founder of Iffel International and author of The AI Translator. Iffel International helps small and mid-sized law firms build AI governance and growth strategy without creating dependency on outside executives.

Here is another article dedicated to Managing Partners of Law Firms on How To Manage A Law Firm In The Age of AI: https://www.iffelinternational.com/how-to-manage-a-law-firm-in-the-age-of-ai/