State AI Ethics Rules for Attorneys: What IP Clients Should Ask in 2026

Yes, you can ask, and now the law is starting to back you up. State AI ethics rules for attorneys are moving from bar guidance into hard statute, and that shift hands IP clients in 2026 disclosure rights they rarely think to use. California is the front-runner. On August 31, 2026, its Assembly passed Senate Bill 574 by a 75-0 vote, and the Senate approved the amendments and sent it to Governor Newsom. If signed, it makes California the first state with statutory rules on how attorneys and arbitrators may use generative AI. For a small-business owner whose patent or trademark work runs through an AI-assisted firm, that is not abstract. It is a checklist you are now entitled to run.

What SB 574 Actually Requires

The bill is narrower than the headlines suggest, and more useful because of it. It sets four obligations. Attorneys must keep nonpublic information out of public AI systems. They must ensure their AI use does not unlawfully discriminate. They must take reasonable steps to verify the accuracy of AI material and correct any hallucinated output. And the filing attorney must personally read and verify every citation, including ones AI did not generate.

Arbitrators get a harder line. Under SB 574 they may not delegate any decision-making to an AI model, may not rely on AI-generated information outside the record without disclosing it to the parties, and must assume responsibility for every aspect of an award regardless of what tool assisted it. If a tool cannot cite an independently verifiable source, the arbitrator may not assume that source exists.

This is the enforcement layer catching up to voluntary standards. ABA Formal Opinion 512, issued July 2024, set the national framework on competence, confidentiality, and supervision. By early 2026, forty-seven state bars had issued their own AI ethics guidance. SB 574 turns that guidance into something a client can point to.

What State AI Ethics Rules Mean for Attorney IP Clients in 2026

The stakes in intellectual property are unusually unforgiving. A patent claim lives or dies on precise language. A single misstated word can alter the scope of the entire claim and leave it vulnerable during litigation, which is exactly the kind of error a fluent, confident AI draft can slip past a rushed reviewer.

The verification duty in SB 574 maps directly onto that risk. If your attorney must verify AI output before it reaches a court, you have every reason to expect the same standard on a patent specification or a trademark clearance opinion that never sees a courtroom. The disclosure duty gives you the second lever. You are not being difficult by asking which tools touched your filing. You are asking your lawyer to document what a new statute already expects.

The Disclosure Checklist to Run With Your IP Attorney

Bring these questions to your next call. The strength of the answers, not their comfort, is the signal.

• Which AI tools touched my matter, and at which step? Drafting, prior-art search, clearance, or intake. A firm that has integrated AI deliberately can name the tool and the task.
• Who personally verified the AI output before it reached me or the USPTO? You want a named attorney and a defined workflow, not “the tool is reliable.”
• Was any of my confidential information entered into a public AI system? Under SB 574 and Model Rule 1.6, nonpublic data does not belong in public tools. Ask about zero-retention vendor agreements.
• Will you tell me in writing when AI generates a material portion of my work product? Proactive disclosure in the engagement letter is the tell of a firm that takes this seriously.

For the deeper version of this conversation, our seven questions to ask a law firm about AI before you hire breaks down what a strong answer sounds like versus a marketing answer. And because a confident wrong result is the hardest error to catch, read how an AI hallucination in a trademark or patent clearance search can fake a clean result that a real search would have flagged.

When AI-Assisted Patent Work Contains Errors, Who Pays

This is where clients tend to stop asking, and where the money is. Through 2024 and 2025, almost every US firm carried what insurers now call silent AI cover: a standard professional-liability policy never amended for AI, applied to an AI claim after the fact. That silence is ending. By late 2025, major legal-malpractice carriers including CNA and AXA XL had begun revising policy language, and seven of thirteen surveyed carriers reported a rise in AI-related claims.

The claims are real, not hypothetical. AI liability tracker Testudo has logged 720 US incidents where an attorney was sanctioned or disciplined over AI-hallucinated content, with the highest recorded fine reaching $110,000 as of 2026. When an AI tool contributes to a malpractice event, the claim can land in a gray zone where neither the professional-liability policy nor the cyber policy clearly responds. So ask a plain question. If AI-assisted work on my patent or trademark contains an error that costs me a right, is that claim covered under your current policy, or excluded? A firm that has read its own renewal terms will answer without flinching.

Your recourse also depends on who is accountable for the final product. The lawyer signing your filing owns it, AI or not, which is the same principle SB 574 writes into arbitration. If you want the fuller picture of what oversight you can demand, our guide to attorney authority, scope, and your oversight rights covers where a client’s say begins.

The Shift Worth Watching

California will not be the last state here. SB 574 gives every other legislature a template, and it gives you a standard to hold your counsel to even before your state adopts one. The point is not to fear AI in your IP work. Used well, it lowers cost and speeds drafting. The point is that a good attorney will welcome these questions, because answering them is now part of the job.

If you are choosing IP counsel this year, or unsure what your current firm actually does behind the screen, a short conversation with a lawyer who can walk you through your AI-disclosure rights is worth more than another vendor brochure.