California legislature passes first-in-the-nation rules on how lawyers and arbitrators use generative AI
California's Legislature unanimously passed Senate Bill 574 on August 31, 2026. As summarized by Sullivan & Cromwell, the bill would bar attorneys from delegating the practice of law to generative AI, limit entering confidential or nonpublic information into AI systems unless access is restricted to authorized, confidentiality-bound people, require lawyers to verify AI output and citations, and require disclosure of AI use in court filings. It would also reach arbitrators, judicial officers, and dispute-resolution providers. As of late September the bill awaited the governor's action, so it is not yet law.
California just passed a bill, unanimously, that spells out what lawyers can and can't feed into an AI tool — and your confidential information is first on the list. If your outside counsel or other advisers use generative AI on your matters, their AI habits are now part of your own risk picture, and most companies have never asked.
Post or short talk segment: "You vetted your AI vendors. Did you vet your lawyers' AI?" Vendor risk extends beyond software to every professional-services relationship that touches company data. The failure point is rarely the tool; it's whether anyone defined who may use it, on what, and with what safeguards — the human layer again.
Source: Source: Sullivan & Cromwell via CLS Blue Sky Blog, Sep 28, 2026 · clsbluesky.law.columbia.edu