California Bar Reveals It Used AI For Exam Questions, Because Of Course It Did
The hits just keep on coming. The post California Bar Reveals It Used AI For Exam Questions, Because Of Course It Did appeared first on Above the Law.


Unaware that you’re supposed to rip the band-aid off all at once, the embattled officials behind the California Bar Exam decided they hadn’t had enough of the non-stop cavalcade of disastrous headlines about February’s exam, so they just casually threw out there that they used AI to come up with some of the questions.
Everyone’s going to be chill about this, right? I mean, “bar exams” and “artificial intelligence” are two topics that famously elicit only only rational, measured reactions.
After February’s haphazard glitch-fest was the product of a shotgun wedding revamp that ditched the NCBE in an effort to save the state licensing operation from descending into bankruptcy. The NCBE’s testing venue rules had pushed California’s resources to the limit with the most populous state in the union forced to book massive, expensive, and mostly inconvenient locations every time it offered the test. A new provider would offer the examiners the opportunity to do more remote testing and take advantage of multiple, smaller locations to save money. It sounded good on paper and, more or less, it’s still a better path forward for California.
But things didn’t quite work out this time.
They commissioned Kaplan to write the questions and Meazure to administer the test. Then there were practice test problems, technical problems galore, the more convenient venues didn’t materialize, and the proctoring spawned subreddits worth of horror stories. They ended up having to offer refunds and make-up tests before floating the possibility of inviting more chaos by throwing the whole project in the can.
Making this the perfect time to inform the not-at-all-stressed applicants that the questions might have also been hallucinated AI slop. THANKS, BYE!
The State Bar of California said in a news release Monday that it will ask the California Supreme Court to adjust test scores for those who took its February bar exam.
But it declined to acknowledge significant problems with its multiple-choice questions — even as it revealed that a subset of questions were recycled from a first-year law student exam, while others were developed with the assistance of AI by ACS Ventures, the State Bar’s independent psychometrician.
Despite what it sounds like, psychometricians are not villains from an L. Ron Hubbard book. But they also are not lawyers, understandably troubling educators and applicants when they learned that a bunch of non-lawyers used AI to develop questions. Basically, their job is to measure test performance, not use ChatGPT to rewrite “Contracts for Dummies.”
Before rushing to rage, the AI-aided questions made up a “small subset” of the exam, amounting to 23 of the 171 scored multiple-choice questions. It’s also worth noting that developing questions isn’t the same as handing over test design to the computers. The chair of the State Bar’s Committee of Bar Examiners, Alex Chan, pushed back against that prospect, explaining, “the professors are suggesting that we used AI to draft all of the multiple choice questions, as opposed to using AI to vet them.”
Over and above whatever it means for AI to “vet” them, the bar also said that all questions were reviewed by subject matter experts. Of course, those same panels suspiciously lost a few credentialed law professors after the Bar worried that academics who worked with the NCBE in the past could raise copyright issues. Theoretically the non-profit NCBE could waive those in the public interest, but it’s also a non-profit with $175 million in assets so make your own conclusions about how that would’ve gone down.
There’s nothing inherently wrong with AI helping out in this process. If the examiners have confidence in the subject matter experts reviewing the final product, it might be a useful brainstorming tool. Indeed, the AI doesn’t seem as concerning as who allegedly used it. Garbage in, garbage out is a real problem, and whether it’s writing or vetting the questions, having a non-lawyer on the other end of the keyboard raises risks. Some of the professors cited in the Times article also worry that it raises conflict of interest issues as the psychometrician is expected to vouch for the reliability of the questions on the back end, though the bar examiners stress that this process isn’t subjective.
Besides, the bar examiners were told to consider using AI by no less than the California Supreme Court — who oversees the process. A statement that prompted the California Supreme Court to immediately respond that they never knew anything about this until the press release dropped.
WELL. OILED. MACHINE.
There are a lot of risks in using AI generally and to build a life-changing exam specifically. Other than the inherent shadiness involved in making this announcement a couple months after the fact and having the ultimate authorities at the State Supreme Court reply, “Whoa! Don’t rope us into this mess!” there’s reason to believe the bar examiners probably used AI responsibly here.
Still, Katie Moran, an associate professor at the University of San Francisco School of Law, told the Times that the State Bar should “release all 200 questions that were on the test for transparency and to allow future test takers a chance to get used to the different questions.” That’s the sort of proactive transparency that avoids these sorts of belated bombshells. It also puts more eyes on the questions and helps the authors — the human ones — refine questions in response to the wisdom of the subject matter expert crowd.
“She also called on the State Bar to return to the multi-state bar exam for the July exams.” Yeah… for the love of all that’s holy, don’t do that. Nothing solved by more whiplash, there’s almost no chance the examiners could lock down all the pricey venues they would need by July, and — the State Bar notes — one of the few things applicants hate more than this test is the lack of a remote option.
Also, the NCBE questions consistently generate a ton of applicant complaints. While everyone in California — rightly — grumbles about these new questions, just wait until July when we start seeing the metric shit ton of rage over incoherent multi-state questions. Eventually… the NCBE bans talking about the substance of the exam, ostensibly to protect the sanctity of the exam for later test takers but with the useful side effect of keeping criticism tamped down until after results are released. TRANSPARENCY!
The California Bar Exam is broken in a lot of ways and people should be angry about it. But without a time machine to go back a number of years to when they should’ve started planning to move to this new test instead of slapping it together in a matter of months, it is what it is. But they’re working on fixing it, even if this was a borderline irresponsible move to start this in February.
And of all the problems we’ve heard throughout this process, using AI to develop some questions doesn’t make the top-tier.
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search.
The post California Bar Reveals It Used AI For Exam Questions, Because Of Course It Did appeared first on Above the Law.