THE POLICY

For Professor Madison’s courses, seminars, and all other graded work, these are actual rules – not simply guidelines:

Each piece of work product that a student submits – a memo, email, slide deck, or other material for a course on intellectual property law; a portion of a research paper for a seminar or independent study project (summary, outline, draft, final version) must be accompanied by a completed “AI Disclosure” form that either (i) uses this form (the link goes to a downloadable MSWord document for students to complete, save, and submit or (ii) using its own form and format answers the following questions:

AI Disclosure
[Course name and semester, year]

Student name
Date of this disclosure

[As appropriate]
Assignment [One] [Two] [etc] or Paper [Initial Summary] [Outline] [etc]

Choose and proceed accordingly

Completing that form goes hand-in-glove with the balance of the policy, which is this:

All work that a student submits for a grade will bear their name, which means that if any of it originated with a machine, a robot, a chatbot, or an AI – whether in response to prompts from you or otherwise – then the student and only the student is responsible and accountable for its accuracy, clarity, consistency, persuasiveness, and responsiveness. Blaming the AI is not permitted. Failing to disclose AI use per the above form is not permitted. Either action will result in the student’s being referred to the appropriate Dean for disciplinary review.

RATIONALE AND EXPLANATION

Generative AI systems have so rapidly and comprehensively become part of the world of law practice and other portions of the legal profession and the legal system that it makes no sense, at least to me, to limit their use by people training to become members of the legal world and its relatives.

Likewise, my anecdotal observation is that essentially all students entering law school since 2024 (more or less) have worked or played with Generative AI systems to a significant degree in their undergraduate programs, other graduate programs, in jobs, and/or just for fun. It makes no sense, at least to me, to tell those people (you) to “unring the bell.” As someone whose responsibility goes mostly to training law students to become lawyers, I should no more try to prevent them from using Generative AI than I should prevent them from using Westlaw or LexisNexis. I should not require that students only look up cases in physical reporters. I should not require that students ignore headnotes, or that they use desktop computers rather than tablets or laptops.

That said, any technological system in any setting shapes how people act and think in lots of ways, only some of which are typically salient and visible to “users.” I have listed some of those below, not because I think that they are universal or even common but because public commentary around AI systems have highlighted them.

So, while I do not expect or require that students use AI systems or platforms or agents et cetera, and I do not forbid students from using any of those things, I do intend that students should be (or should become) increasingly aware of the virtues and drawbacks of their choices. The policy summarized above is aimed at requiring students to be explicit about those choices and, I hope, to learn something about themselves and their learning processes along the way.

In the best of educational worlds, creating opportunities for self-awareness when it comes to technological systems (call it “critical AI awareness,” for example) should be paired with creating systems of thoughtful instruction around the origins, uses, benefits, and risks of technological systems. My courses cannot provide that thoughtful instruction at the same time that they aim to teach the human skills that the courses focus on. One hopes that Pitt Law (and other law schools) will eventually move in the direction of providing thoroughgoing teaching around law and technology, in practice.

What Generative AI Systems Likely Can – and Cannot – Do

Origins, uses, benefits, and risks?

Since ChatGPT was released publicly in late 2022, most people have learned that AI chatbots can be useful – to a point – in producing text that is factually accurate with respect to summarizing bodies of material, and in producing text that is “administrative” or “procedural” in character. An AI can write a manager’s memo to employees; an AI can summarize a book. But AI systems can make mistakes (“hallucinate,” as the saying goes, and more on that below), and an AI is likely to do a poor job of exercising what we usually think of as human judgment (“is this a wise thing to do? why or why not?”) or evaluating questions of value(s) or ethics.

Today, chatbots are only the tips of the proverbial icebergs (plural) when it comes to the capabilities and uses of Generative AI systems (and other AI systems), in law or otherwise. This page is not the place for me to elaborate. For anyone interested in thoughtful “ordinary” uses of bots, agents, and more, follow Ethan Mollick at the Wharton School (about AI capabilities generally) and Seth Chandler at the University of Houston Law Center (about AI capabilities for lawyers and law students. Both popular literature and scientific literature on the large-scale risks of Generative AI, particularly of so-called “frontier” models, is easy to find. So is both popular and scientific literature on the abundant virtues of Generative AI and other AI systems in certain sectors, particularly in health and medicine, and in environmental policy, and in cybersecurity.

The richness and complexity of the AI landscape means that simple policies that ban AI altogether (or try to, or claim to as a matter of signaling “values” even while the policy is unenforceable), and AI “strategies” that try to put “AI training” in a single box (such as: AI is a tool for students to get familiar with as part of their research and writing education) are likely just that … simple solutions for a complex world. Compare recent “AI for law students” policies adopted by law schools at UC Berkeley and the University of Chicago, each of which is, in my view, more restrictive than is wise for contemporary law students and each of which will, soon enough, come under pressure for revision as AI use evolves in practice. The recent announcement of an AI strategy by the law school at the University of Texas draws a better balance, all things considered.

The deans and faculties of those law schools have legitimate concerns. As with all law and policy development, there is more art than science in translating those concerns into appropriate guides for action. A complex landscape requires careful, layered, nuanced treatment. Consider the combination of questions raised below.

For my students, the concerns for you to focus on include these:

[1] Don’t let the AI do your thinking for you.

What some critics characterize as “cognitive offloading” – the phenomenon of humans relying on machines to do “their” thinking, rather than engaging in human intellectual engagement, is an example of a possible problem. “Cognitive offloading” is not uniquely associated with use of Generative AI systems; lawyers have engaged in “offloading” for decades by, among other things, using West headnotes to guide their legal research. Academic settings that might try to ban AI use, including some law schools and law professors, claim to be trying to get students to train to think “authentically,” in the manner of pre-technologically-literate lawyers, before being introduced to AI systems. I have written elsewhere about the deep flaws in the premise that legal education ever plausibly claimed to train people to “think like lawyers.” To act as lawyers, write as lawyers, and speak as lawyers … yes. Law schools do that, or they should. “Think” “like” lawyers? No.

[2] Watch out for fake facts.

“Hallucination” is the name that society has adopted as a label for the phenomenon of AI systems generating information (such as case or statutory citations) that is factually untrue or even nonexistent. The label is a terrible one, because it indulges the mistake of believing that AI systems are “thinking.” Avoiding hallucination is easy, in principle, if time-consuming: make sure that every case and statute that you cite, and every quotation that you use, is traceable to an authentic primary source.

Similarly, AI systems might “hallucinate” facts and events and make up images and other visual material.

[3] Ethical landmines are everywhere.

In the context of my courses, the fact that I give students permission to use AI systems means that ethical concerns are reduced (but see [1] and [2] above, which are ethical issues). In a non-academic setting – a summer job, an internship or externship, and any paid work, during or after law school, it is important to be aware of (i) how any AI system that you might use will (or will not) capture and re-use information that you feed to it, via prompting or otherwise (this is largely but not entirely a privacy issue); (ii) how any AI system that you might use will (or will not) interact with other IT systems inside or outside “your” organization (this is largely but not entirely a confidentiality issue); (iii) how any AI system that you might use was built, using source or “training” data (this is partly a confidentiality issue and partly an intellectual property law issue); and (iv) the nominal and true costs of any AI system that you might use (as a student and as a “free” user, these systems cost money to run, even if you – as a student – do not pay anything, or much).

“Free to use” chatbots and other AI systems usually have more privacy and confidentiality invasive structures than pay-to-use chatbots and AI systems do.

[4] Don’t lose your own “voice.”

Once you have read a lot of AI-generated text (which I have: I subscribe to a lot of Substacks, where Anthropic’s Claude system has a lot of admirers), then it is difficult to “un-see” text written by AI. Even though I have no general concerns about AI use, I tend to discount AI-written text. Not because it is wrong or misleading, but instead because it is dull. Even lawyers (and so on) need to learn to write about law and clients in their own “voice,” which means with their own style, syntax, and vocabulary. Every law student reads opinions by famous judges – Hand (Learned, and brother Augustus), Cardozo, Holmes, Brandeis, Frankfurter, Scalia. What made them famous, in each case, was the judge’s distinctive blend of wit, style, and substance. Developing that blend takes years. AI threatens to cut it off. Who wants a world – a legal world, or any other world – where everyone sounds the same?

That advice comes with an important caveat. Sometimes, professionals work in organizational settings where your voice is literally unimportant. As an individual, you are expected to speak in the voice of the organization, and/or in the voice of the field, discipline, or business/government/nonprofit sector of which it is a part, and/or even in the voice of a supervisor. Sometimes, you are expected to speak or write in a voice that represents combinations of those things.

What to do?

First: learn to see and to understand your role. Are you speaking for yourself? Or are you speaking as a human embodiment of something or someone else? Figuring out the difference may be difficult, especially early in your career. It requires observation, curiosity, and experience.

Second: learn the different combinations of AI virtue and AI vice in those different settings. An AI system may help you speak and write in an organizational voice even if it undermines your own individual voice. But privacy, confidentiality, security, and other concerns identified above will still apply, and they may even apply with greater force in an organizational context than in what seems to be a purely personal one.

To Sum Up

If you are tempted to explore the capabilities of ChatGPT or other AI-powered chatbots and systems in the course of writing your papers, feel free to do that. Explore. Learn how to use prompts effectively. Play with chatbots as you wish. You might use AI systems to generate ideas, to develop resources, and to refine your writing.

In some respects, AI systems are unavoidable in legal education, because Westlaw, Lexis/Nexis, and other legal research services now include AI elements in their research platforms.

If you use a writing service such as Grammarly, know that those services, too, include AI elements.

Importantly, as a new lawyer you are likely to be expected to use AI systems effectively and efficiently to support both your clients and your colleagues. AI is rapidly become part of law, rather than simply another tool in a lawyer’s toolkit.

As a general rule, all of this is essentially the same guidance that you will be expected to follow in practicing law. Here as in other aspects of this course, my guidance and requirements are meant to emulate those that you will encounter in your professional careers.

All I ask is that you think about what you are doing and that you take the time to write it up and submit your summary when you submit your work for a grade.

If you have any questions or concerns about what that policy means, consult me before turning in your work.

IF YOU WANT TO USE AN AI SYSTEM, WHICH ONE(S) SHOULD YOU CHOOSE AND WHY?

I have only a little advice for students who want guidance as to which AI systems to use in law school, and when, and why. That advice is:

Read useful advice from experts who have done the research and have thoughtful recommendations.

Here are three pieces of useful advice:

Professor Seth Chandler, University of Houston Law Center, “What AI should I get for law school? 2026 edition“

Professor Ethan Mollick, Wharton School, University of Pennsylvania, “An opinionated guide to which AI to use to do stuff“

Professor Kevin Frazier, University of Texas School of Law, “Frazier on AI for non-tech students“

ACKNOWLEDGEMENT

The policy and the disclosure form are borrowed, with gratitude and modifications, from Professor Madelyn Sanfilippo at the University of Illinois School of Information Sciences.

This page was last updated in August 2026.

.