Our Take
Banning tools students can freely access anywhere teaches them to hide AI use, not to think critically about it; disclosure creates accountability and forces the harder pedagogical work.
Why it matters
Law schools are the training ground for how legal professionals will eventually use AI in practice. A policy that drives tool use underground rather than into the open wastes an opportunity to teach judgment and verification skills before graduates enter firms that will demand both.
Do this week
Law school administrators: audit your AI policy against Bhargava's three-part framework (use-based tasks only, mandatory disclosure, concrete safeguards) before next semester so students learn to work with AI rather than around your rules.
Berkeley's ban meets resistance from within
Last month, Berkeley Law School prohibited AI use across broad areas of study, citing concerns about intellectual development. The school wanted to protect students from over-dependence on language models.
Lakshita Bhargava, a Master of Laws student specializing in AI Governance at Berkeley, pushed back publicly. She acknowledged the school's intent but argued the approach was fundamentally flawed. "A permitted-but-disclosed model creates better accountability and feeds into learning," she told Artificial Lawyer. Bhargava recently worked at Shardul Amarchand Mangaldas, a leading Indian law firm.
Her critique centers on a simple fact: students have free access to AI tools outside the classroom. A blanket ban does not eliminate use. It only eliminates visibility.
Hidden AI use is worse than transparent use
Bhargava outlined a three-part alternative that reframes the institutional problem. First, teach students to deploy AI for work it does well (summarizing, outlining, editing, translating, study prep) and nowhere else. This creates a task-based boundary, not a moral one. Second, require disclosure of any AI use, so professors know what actually happened and can discuss it. Third, set concrete guardrails: no confidential data, mandatory citation verification, full student accountability for final work.
The logic is sound. If a student uses AI to brainstorm but knows disclosure is forbidden, they will either hide the work or use AI without discussing it with anyone. Neither outcome teaches judgment. If disclosure is expected and discussed, the student and professor together examine where the tool helped and where it failed.
As Artificial Lawyer noted in its editorial addition, banning AI in education mirrors banning books from libraries. The books still circulate; they circulate invisibly. Better to widen students' minds so they can assess what they consume, whether from AI or otherwise.
Dr. Megan Ma of Stanford Law, cited in the piece, offered a further point: AI can deepen legal understanding by surfacing counterarguments a student might miss. Prohibition forecloses that possibility. Transparency unlocks it.
The policy question law schools must solve
Berkeley's ban reflects genuine anxiety about AI's role in education. But the response treats the symptom (students relying on AI) rather than the disease (students not learning to verify, question, or integrate AI output into their thinking).
The harder work is Bhargava's framework: defining the tasks where AI helps without replacing skill-building, requiring students to name their use, and holding them accountable for the result. This demands faculty engagement. It cannot be automated or delegated to an honor code.
Schools that pursue this path will train lawyers who know when to use AI and how to catch its errors. Schools that ban AI will train lawyers who use AI and hide it. The gap will matter most to clients who do not know which they hired.