California lawmakers are moving to force one of the state’s most powerful licensing bodies to admit, on the record, when it lets artificial intelligence help write the test that decides who gets to practice law. Assembly Bill 1651, authored by Assemblymember Dixon, cleared the full Assembly 68-0 in April and the full Senate 36-0 in June, and now sits before the Assembly for concurrence ahead of the legislature’s August 31 adjournment for the year.
The bill is a direct response to the February 2025 bar exam debacle, still fresh in the memory of California’s legal community. That exam, the first developed by a private contractor after the State Bar moved to a hybrid remote format to close an $8 million deficit in its lawyer admission fund, was plagued by technical failures so severe that nearly 90 percent of test takers were later offered the chance to retake it. UC Berkeley Law Dean Erwin Chemerinsky called it a display of “stunning incompetence from an entity that exists to measure competence.”
The deeper scandal, though, was what came out afterward: the State Bar acknowledged that 23 of the exam’s 171 scored multiple-choice questions had been developed by a firm called ACS Ventures with the assistance of AI, and that another 48 were recycled from an older first-year law students’ exam. Only 100 questions came from Kaplan, the contractor the Bar had actually hired to write the test. The State Bar has defended the results, saying all questions passed through content-validation panels and subject-matter experts, and that the combined multiple-choice section scored a 0.89 reliability rating, above its own 0.80 psychometric target. That defense has not satisfied critics. UC Irvine Law’s Mary Basick called the lack of disclosure “an egregious breach of trust,” and noted the plan had not, to her knowledge, been approved by the California Supreme Court, which oversees bar admission.
AB 1651 does not ban the State Bar from using AI to help write exam questions or study materials. What it requires, starting January 1, 2028, is disclosure: the State Bar must post on its website when AI-generated content was used in developing or administering its examinations, and it must print a disclosure on the cover of any study materials it publishes or endorses that contain AI-generated content. Critically, the bill’s own analysis notes the duty applies only to content developed by or at the explicit direction of the State Bar itself, and the disclosure requirement holds regardless of whether a human later reviewed or revised that AI-generated content. The bill carries no fiscal impact, has no registered opposition, and is sponsored by the California Civil Liberties Advocacy group.
The measure is one of roughly two dozen AI-related bills the legislature is racing to finish before its session ends. Others moving in parallel include SB 928, which would bar California State University campuses from using AI in place of human instructors and has already reached Governor Newsom’s desk after passing the Assembly 37-0. Taken together, this session’s AI bills reflect a broader legislative mood: less an attempt to restrict what AI can do inside California’s public institutions, and more an effort to guarantee that residents know when it has been used at all, on tests, in classrooms, or in the material meant to prepare them for either.
That distinction is the one worth sitting with here at The Navarro Report. This publication operates under a Human-Directed AI Journalism model, disclosed on every piece we publish, precisely because we believe the public deserves to know when AI assisted in producing content that shapes their understanding of civic life — the same principle AB 1651 would write into law for the body that licenses California’s attorneys. The State Bar’s February 2025 exam failed not because AI touched the process, but because no one said so until forced to. A licensing exam and a news article are different kinds of institutions, but the underlying obligation is the same: when a machine helps build something the public is asked to trust, the public gets to know.
If AB 1651 clears its final Assembly concurrence vote and reaches Governor Newsom’s desk before adjournment, California would become one of the first states to write AI-disclosure requirements directly into the statute governing attorney licensure — a modest step, but one that treats transparency as a baseline expectation rather than a courtesy. For the thousands of applicants who sit for the California bar exam each cycle, and for the law schools that spend years preparing them, that baseline matters: it is the difference between trusting a credentialing process by default and trusting it because the institution running it has agreed to show its work.
Human-Directed AI Journalism: This article was drafted with AI assistance under direct human editorial direction, fact-checked, and approved for publication by The Navarro Report.