We asked a very new and apparently the most up-to-date AI model a legal question recently, just to see what would happen.
The question was simple: Under the California Rules of Professional Conduct, do attorneys have a duty to report other attorneys for misconduct? The AI answered confidently. It was wrong.
We pushed back. We told the AI we believed the answer was wrong. The AI then searched, found the actual answer, and admitted the mistake outright: California adopted Rule of Professional Conduct 8.3, effective August 1, 2023, creating exactly the mandatory reporting duty the AI had just told us did not exist. The correction was fast, polite, and thorough. It was also too late to matter, because we already had a wrong answer sitting in front of us, stated with total confidence, by a tool that had no idea it was wrong until we told it so.
That is the story we want every applicant studying for the California Bar Examination to sit with for a moment.
In over 25 years of preparing applicants for this exam, we have watched technology after technology promise to make studying easier. The internet. Online outline banks. Now artificial intelligence. Every wave arrives with the same pitch, and every time, we end up saying the same thing: easier is not the same as accurate, and it is certainly not the same as prepared.
Here is what makes the AI problem different from the outline-bank problem of years past. A bad outline is obviously a bad outline. You can sense when something feels thin. A confidently wrong answer from an AI tool does not feel wrong. It feels like an answer. It is fluent, organized, and stated without hesitation, whether the underlying rule is current or three years out of date. That confidence is exactly what makes it dangerous in a way that a stale textbook never was.
Rule 8.3 itself is instructive here, and not just as a cautionary anecdote. It requires a lawyer who knows of credible evidence that another lawyer has committed a criminal act, or engaged in conduct involving dishonesty, fraud, deceit, or misappropriation, to report that lawyer to the State Bar or to a tribunal with jurisdiction to act. California rejected an equivalent rule twice before finally adopting it in 2023, in significant part because of the fallout from a single high-profile attorney misconduct scandal. This is precisely the kind of rule that shifts. It is precisely the kind of rule an AI tool trained on older data will get wrong with complete confidence. And it is precisely the kind of rule the bar examiners expect you to know cold, current as of today, not current as of whenever a model happened to stop learning.
We want to be fair here, because the point of this piece is not that AI is useless. It is not. Used the right way, AI can be a reasonable supporting tool. It can help you generate a first draft of a flashcard set you intend to rewrite yourself. It can offer an alternate explanation of a concept you are already studying through primary sources. It can help you organize your own notes once you have already done the thinking. Used that way, AI is an accompaniment.
The danger is when AI stops being an accompaniment and becomes the source. When you ask it a substantive question about California law and simply accept the answer, you have no way of knowing, in that moment, whether you have received Rule 8.3 in 2026 or the rejected proposal from a decade earlier. You will not find out you were wrong until someone corrects you, the way we corrected the AI in our example. On exam day, there is no one there to correct you. There is the fact pattern, the call of the question, and whatever you actually know.
This is also not a theoretical concern confined to obscure professional responsibility rules. The California State Bar’s own use of AI to help generate exam questions in 2025 produced widespread complaints about inaccuracies and phrasing that did not track standard legal terminology. If AI can introduce errors into the exam itself, with actual bar examiners reviewing the output, it can certainly introduce errors into your outline when no one is reviewing it but you.
COPRAC has already weighed in on this for licensed attorneys, and the guidance translates directly to bar applicants. Attorneys have been sanctioned for filing AI-generated briefs containing fabricated case citations. The guidance is clear that an attorney’s lack of awareness of AI’s limitations does not excuse the inaccuracy. Read that again as an applicant. If “I did not know the AI was wrong” is not a defense for a licensed attorney standing in front of a judge, it is certainly not a strategy for an applicant standing in front of the California Bar Examiners.
So if you are using AI in your preparation, here is the standard we want you to hold it to, the same standard we held it to in our own example. Never accept a substantive answer about California law on the first response. Ask the tool to verify against a current, authoritative source before you write it on a flashcard or build an outline around it. If a rule sounds like it might have shifted in the last few years, treat that as a reason for more scrutiny, not less. And do the underlying work yourself: the organizing, the connecting, the wrestling with exceptions, because that process is where retention actually comes from. An AI tool can hand you a polished answer in three seconds. It cannot hand you the understanding that lets you spot the issue inside a messy fact pattern, apply the current rule correctly, and defend your reasoning in an essay. That skill is built one way. You build it.
We will say plainly what we already know some of you are thinking: yes, we used AI to help with parts of this very piece, and yes, we caught it making exactly the kind of mistake we are warning you about. We are not above the conversation we are asking you to have with yourself. We checked the work. That is the whole point.
AI is a tool. In the right hands, used the right way, with everything it tells you checked and rechecked, it can support your preparation. It cannot replace your preparation. The exam will not ask the AI. It will ask you.
For more information, or to schedule a confidential consultation please contact execbar.com.
888.393.2392
Contact Us
Frequently Asked Questions
Can California Bar Exam applicants use AI to study?
Yes, but AI should be used only as a supporting tool. Applicants should verify substantive legal answers against current, authoritative sources before relying on them.
Why is AI risky for California Bar Exam preparation?
AI can give fluent, confident answers that are outdated or wrong. That is especially risky when California law has changed recently, such as with Rule 8.3.
What is California Rule of Professional Conduct 8.3?
Rule 8.3 is California's mandatory reporting rule for certain attorney misconduct. It became effective on August 1, 2023.
Should applicants rely on AI-generated outlines or flashcards?
Applicants should not rely on AI-generated outlines or flashcards unless they have checked the rules against primary or authoritative sources and rewritten the material themselves.
What is the safest way to use AI while studying for the California Bar Exam?
The safest approach is to use AI for organization, alternate explanations, and drafting study aids, while treating all substantive statements about California law as unverified until checked.