Should California’s AI law govern lawyers, disclosure will not erase review

California Governor Gavin Newsom signed into law on October first legislation setting rules for how lawyers may use generative artificial intelligence in their work, Reuters Legal reports. Bloomberg Law, Business & Practice reported the signing under the headline Newsom Signs First-of-Its-Kind Bill on Lawyer, Arbitrator AI Use. The California measure is part of a slate of AI-related measures signed by Newsom, according to Reuters Legal. Legaltech News (Law.com) reports that SB 574 also bars arbitrators from relinquishing their decision-making powers to generative artificial intelligence.

The law moves the accountability question from internal practice to an express California rule. Bloomberg Law says California’s first-in-the-nation law gives law firms the task of determining when their use of generative artificial intelligence requires disclosure. That matters for an AI-native legal business because finished legal work remains exposed to review even when the work is produced with automated assistance. Disclosure can identify the use of the technology; it does not make an inaccurate filing accurate, restore a missed obligation, or transfer responsibility to the software. The statute therefore supports a narrow but important proposition: California has made AI use part of the conditions under which legal work is delivered and assessed. It does not support the broader proposition that regulation makes AI-native work reliable.

The event proves that California has enacted a framework governing lawyers’ use of generative artificial intelligence and that the framework includes disclosure decisions and a limit on arbitrators’ delegation of decision-making. It does not prove that lawyers will disclose use consistently, that clients will receive better work, or that AI-native firms will avoid sanctions or other consequences for defective work. Legaltech News (Law.com) supplies the clearest boundary: its report says SB 574 bars arbitrators from relinquishing their decision-making powers to generative artificial intelligence. That rule addresses who may make the decision; it does not establish that every AI-assisted legal product has been checked, or that human review has caught every error. A stronger reading would require a California order, enforcement action, or completed matter showing how the law treats defective AI-assisted work. No such result appears here. The enacted rule creates accountability conditions; it does not demonstrate compliance or performance.

The next fact to watch is a California disclosure decision applying the law to a specific use of generative artificial intelligence by a law firm, followed by any California order or enforcement action addressing the quality of the resulting legal work. A decision that treats disclosure as separate from the duty to review would confirm the narrower reading: notice does not replace verification. An order imposing consequences for an AI-assisted error would make the position concrete. Conversely, a California ruling that treats disclosure as sufficient protection without examining the work would cut against it. We expect the important test to be observable in the treatment of the finished matter, not in the existence of the technology or the firm’s description of its process.