If AI enters legal work, California puts review on lawyers

Newsom signed SB 574 in California, enacting a bill on lawyer and arbitrator use of artificial intelligence, Bloomberg Law, Business & Practice reports. Legaltech News (Law.com) describes the measure as a new California law that gives lawyers stronger responsibilities to review AI work. The law also addresses arbitrators: Legaltech News (Law.com) reports that SB 574 bars them from relinquishing their decision-making powers to generative artificial intelligence. The event is therefore a state action with two defined features: lawyers remain responsible for reviewing AI-assisted legal work, and arbitrators cannot hand their decisions to a generative system.

That matters because AI does not eliminate the consequences of bad legal work. California’s approach keeps responsibility attached to the professional who delivers or decides, rather than treating the use of a generative system as a substitute for review or judgment. The distinction is important for AI-native legal models. A system can produce work, but the California law described by Legaltech News (Law.com) still places stronger review responsibilities on lawyers. For arbitration, the line is sharper: decision-making powers cannot be relinquished to generative artificial intelligence. Bloomberg Law, Business & Practice separately frames the law as creating tough disclosure decisions for California lawyers. Together, those provisions move the legal system toward accountability for how AI work is checked and presented, not merely toward permission to use the technology.

What the event proves is narrower than a general verdict on AI legal work. It establishes that Newsom signed a California bill addressing lawyer review, arbitrator decision-making, and disclosure decisions. It does not establish that AI-generated legal work is accurate, inaccurate, cheaper, or better for clients. It does not establish that every lawyer will review AI output effectively, or that every arbitrator will preserve meaningful human judgment. Those stronger readings would require an observable result from the law: a California order applying the review responsibility to particular work, a disclosure ruling, or a decision involving an arbitrator’s use of generative artificial intelligence. Until then, the law sets consequences and responsibilities around AI use; it does not show how well those controls work in practice.

The next fact to watch is a California order applying SB 574 to a lawyer’s AI-assisted work or to an arbitrator’s attempt to delegate decision-making. Such an order would show whether the review responsibility and the bar on relinquishing decision-making powers operate as real constraints rather than as statutory language. A disclosure decision would supply a second test, because Bloomberg Law, Business & Practice identifies disclosure as a difficult issue under California’s AI law. The result that would cut against the position would be a California ruling showing that AI use displaced meaningful review without producing professional consequences. We expect the relevant measure of progress to be accountable legal work, not the mere presence of AI in the workflow.