AI’s growing role in legal work is sharpening a basic question for law schools and courts: can lawyers use these tools without letting them replace independent judgment? The University of Chicago Law School’s new strategy, set to begin this fall, puts that question into the classroom. The issue reaches beyond whether students can operate a chatbot or write a useful prompt. Legal work depends on identifying what a question requires, checking whether an answer is supported, and taking responsibility for the result. Those habits matter whether a lawyer works with an AI system, a colleague, or a stack of documents.
The school plans to pilot a general ban on electronic devices in core first-year courses, with limited exceptions. It also intends to teach students how to use AI effectively and ethically. The approach reflects a tension facing legal education: students need to understand the technology, but first they must build the reasoning skills their work depends on. A device restriction can give students more room to follow arguments and engage directly with course material, while later instruction can address how automated tools fit into legal practice. The balance won’t be settled by a classroom rule alone. It depends on whether students learn when an AI answer is useful, when it needs checking, and when the underlying question calls for their own analysis.
Errors reach court filings
Louisiana court cases show what can go wrong when AI-generated material reaches legal filings without reliable review. More than a dozen incidents involving inaccurate or fabricated content in AI-assisted pleadings have surfaced in state and federal courts since last August, according to reporting by The Times-Picayune and The New Orleans Advocate. A filing is a formal part of a legal case, and its claims and citations can shape how a court understands the arguments before it. When generated material is wrong, the problem is not confined to a faulty draft: someone must catch the error, correct the record, and account for its presence.
In one case, a federal judge sanctioned an attorney after a law clerk flagged a false quotation. The attorney asked Claude to correct it, but the program produced more fabricated material. The judge said it “just made up more stuff.” The episode illustrates a basic limitation of asking a tool to repair an answer whose accuracy is already in doubt: a fluent revision can still fail to establish what a source actually says. The tool’s confident presentation doesn’t transfer the duty to verify, and a second answer isn’t independent confirmation of the first.
Louisiana attorneys have faced fines and sanctions over these errors. One lawyer said a human clerk would double-check future filings. That response points to the immediate operational challenge: AI can assist legal work, but responsibility for verifying its output still rests with people. Review has to mean more than asking another system to check the first one. In legal writing, a citation needs to lead to a real source, and a quotation needs to match that source in context. A person must be able to explain why the material belongs in the filing.
Teaching judgment alongside tools
That’s why the law school’s paired approach carries more weight than a simple debate over screens in class. A restriction in core first-year courses suggests that some foundational learning may benefit from sustained attention to the material and the discussion around it. Teaching AI use, meanwhile, acknowledges that students will encounter these systems in legal work and need a practical understanding of their limits. The two goals can reinforce each other if students learn the underlying methods before relying on automated assistance.
There’s a difference between learning to produce a plausible legal argument and learning to test one. A system may help organize material or suggest language, but students still need to recognize whether an authority supports the point being made, whether a quotation is accurate, and whether an answer addresses the actual issue. Those are not decorative skills added after the technology training. They’re what make review meaningful.
Here's the thing: a device ban can shape the classroom, but it can’t by itself produce careful lawyers. Students need to practice checking claims, tracing sources, and revising arguments when evidence doesn’t support them. They also need to understand why an answer that sounds polished may still be wrong. If you're working in this space, that distinction affects how you assess AI-assisted work: the useful question isn’t only whether a tool saved time, but whether the human user can stand behind the result.
What the court examples mean for practice
The Louisiana cases are a warning about process as much as technology. Similar systems can generate text that looks orderly while containing unsupported details, so a review process has to test the substance rather than the style. That takes time and legal knowledge. If a lawyer can’t verify a cited authority or confirm a quotation against its source, the draft isn’t ready to file, however quickly it was produced.
And yet, treating every use of AI as equivalent would miss the practical question. The source describes errors in AI-assisted pleadings and a failed attempt to correct fabricated material; it doesn’t establish that every task or use will produce the same result. The relevant discipline is to know what a tool has contributed and to check the parts that carry factual or legal weight. That work may be less visible than generating a draft, but it’s where professional judgment shows up.
This is more significant than it looks. The cases make clear that a lawyer’s responsibility doesn’t end when a program supplies the words, and a law school’s responsibility doesn’t end when it introduces students to the technology. The difficult work is building habits that hold up when an answer is persuasive, convenient, and wrong. Especially when it looks finished.
Implications for legal education
The school’s plan is a pilot, so its longer-term effect isn’t established by the announcement alone. Its value will depend on what students learn from both parts of the approach: focused work in foundational courses and explicit instruction in responsible AI use. The design raises a useful question for legal educators: how can students gain fluency with new tools while still learning to do the reasoning that lets them test those tools? There’s no shortcut around that question.
For courts, the reported cases underline the importance of careful review before AI-assisted material enters a filing. For law schools, they offer a concrete reason to teach verification as part of technology use rather than as a separate afterthought. And for students, the lesson is plain: assistance can change how work gets done, but it doesn’t make the resulting argument someone else’s responsibility. The measure of success won’t be whether students use AI often. It’ll be whether they can tell when its output deserves trust, when it needs correction, and when they should set it aside.