4.9 Law School Policy on Generative AI
Note: The Law School adopted a new AI Strategy Statement on July 9, 2026, which reflects its current vision for AI.
The Law School has adopted the following policy, effective Autumn Quarter 2023. It was revised in Autumn Quarter 2025 and Autumn Quarter 2026. This policy may change as circumstances warrant or based on further guidance from the University.
“AI” refers to artificial intelligence technology, including generative and agentic tools. A generative tool, or a feature of a tool, drafts new text, images, figures, code, audio, or video, or makes substantive edits to them. An agentic tool takes actions or completes multi-step tasks on your behalf, including without a prompt for each step. Familiar examples include Claude, ChatGPT, Copilot, and Gemini, but the term also covers open-source models, features embedded in other software, and emerging forms such as multi-agent systems. Use is defined by function, not by product name or ownership entity, and the governing question is what a feature does at the moment you use it.
As it is used in this Section, “AI” does not include tools that merely assist with error correction and search and retrieval of source information. Tools that identify errors or suggest corrections to grammar, spelling, or punctuation do not count as AI for purposes of this policy, even if they incorporate AI internally. A tool that rewrites entire sentences or passages, however, is generative. Tools that find, rank, or extract information from sources, including library databases and literature search, do not count as AI for purposes of this policy. A generated summary or synthesis of retrieved sources does count as AI, however, even when it appears inside a search tool.
A default applies in the absence of an explicit policy for the class set by the instructor.
a) For exams, unless the instructor specifies otherwise, the use of generative AI is absolutely prohibited in any exam.
Commentary: Students must not prompt or engage with generative AI in any way during an exam and (as was always the case) every word of the exam answer must be produced by the student themselves during the exam period. This default prohibition does not extend to the use of generative AI when studying for an exam, so long as it does not violate any other aspect of this policy. For example, while preparing an outline in advance of an exam, a student could ask tools like ChatGPT for the holding of Marbury v. Madison or for a list of arguments for and against judicial review.b) For all work that a student submits for any form of academic credit, AI may not be used, whether with or without attribution, in a way that would constitute academic plagiarism if the AI were a human author whose work was used without attribution.
Commentary: The crux of this policy, which applies in any context in which a student submits work for evaluation, is the application of existing academic integrity policies.
Thus, some but not all uses of AI in preparing work to submit for a class would violate this default policy. Using AI to brainstorm ideas for a paper—akin to asking a search engine (like Google), professors, or professional contacts for suggestions—would not violate this default policy. Nor would using AI for proofreading. The principle here is that if AI is performing a task that other technologies (or people) currently perform, and such uses are not a violation of academic integrity, then those uses of AI do not violate the default policy.
In contrast, using AI to compose part or all of a paper, or copying or paraphrasing output from AI and passing it off as one’s own writing, would violate this default policy. The same principle applies here: if AI is performing a task that other technologies (or people) currently perform, and such uses are a violation of academic integrity, then those uses of AI do violate the default policy.
The default policy sweeps more broadly than a prohibition against academic plagiarism, however. A student cannot avoid limits on the use of AI that would otherwise apply by attributing their work to AI. It is not plagiarism to copy or paraphrase another’s work, so long as one properly cites that source. But under the default policy, the use of AI with or without attribution is compared to the use of a human-created source without attribution. Thus, using AI to compose all or part of a paper, even if the use of generative AI is fully documented in properly placed footnotes, is a violation of the default policy.
2. Instructors have flexibility to set policy for the use of AI in their classes. The Law School sets a default for use of AI, but each instructor can set their own policies.
3. Deviations from the default policy shall be stated in writing in the syllabus posted on Canvas for the class. The goal of this aspect of the policy is to ensure that all parties have advance notice of the policies that apply to them, and that those policies are as clear as practicable.
4. All uses of AI must be consistent with the University’s policies on confidential and personal information. Under University policy governing the use of generative AI (emphasis added):
The use of confidential data with publicly available AI tools is prohibited without prior security and privacy review. This includes personally identifiable employee data, FERPA-covered student data, HIPAA-covered patient data, and may include research that is not yet publicly available. Some grantors, including the National Institutes of Health, have policies prohibiting the use of AI tools in analyzing or reviewing grant applications or proposals. Information shared with publicly available AI tools may expose sensitive information to unauthorized parties or violate data use agreements.
Commentary: Concerns about the sharing of confidential information through the use of AI are most salient in the clinical context, but they are present in any context where you would be including sensitive information in prompts inputted into AI. AI tools can retain and use user input to train their algorithms.
Any information that is private or sensitive should only be shared with AI tools if protection against training are in place. You should use University-provided tools, such as PhoenixAI and Claude Enterprise in this context.