Week 1
Sep 11
Intro
- CS
- Overview of the Generative AI lifecycle
- Law
- How to read a court opinion / pleadings and procedures in a lawsuit
About the Course
Generative AI has raised new questions of law, from data collection and model training to generated output and the harms that can arise.
Courts apply existing legal doctrines to these technologies, yet how those laws govern generative AI remains unsettled. This seminar will cover the legal procedures and analysis of Generative AI products. The focus will mostly be on judicial precedent in property, contract, and tort doctrines and how they manifest during the AI lifecycle: from data collection, to model training, to inference, to generative harms. Additionally, the seminar will examine some recent state and international regulations.
The goal will be to develop the skills to analyze opinions and assess how courts may resolve ongoing and future disputes involving generative AI, both in traditional doctrines and more recent regulations. What is most exciting about legal issues in Generative AI is that much of the law is not yet settled and will be revealed through the next decade of ongoing high-stakes litigation.
The seminar will feature a mix of short lectures reviewing both technical and legal concepts, followed by discussion of readings. The readings will include both foundational case law and current litigation at various stages. Learning activities outside of class will include 4–5 readings per week, a mid-semester assignment, and using LexFactor for a guided discussion of the reading.
AI or tech policy is not an objective of this course, as there are other courses at Brown that cover policy more comprehensively. I encourage you to take those courses if you are interested in making changes to AI or tech governance and regulations. Additionally, this course focuses more on the new legal issues that arise from generative AI, and less on issues such as bias, discrimination, fairness, and privacy, each of which is better represented in other courses at Brown that focus on those topics specifically.
Learning Objectives
- Read, brief, and critique judicial opinions and apply their reasoning to novel AI fact patterns.
- Trace how copyright, contract, and tort doctrines map onto data collection, model training, and generative outputs.
- Articulate the fair-use analysis and distinguish the separate copyright questions raised by training versus output.
- Evaluate when AI-caused harm gives rise to liability, and whether AI output should receive First Amendment or Section 230 protection.
- Connect the relevant computer-science mechanics to the legal questions they create: scraping, copying, tokenization, prompting, and retrieval-augmented generation.
Schedule
Readings are subject to change as the seminar progresses. Please do not read ahead more than one week.
Week 2
Sep 18
Acquiring data through scraping
- CS
- Scraping the web, authorization, robots.txt
- Law
- Terms of service, contracts of adhesion, breaches, CFAA
- Cases
- hiQ v. LinkedIn (two court outcomes), Meta v. Bright Data, Register.com v. Verio, Reddit v. Anthropic
Week 3
Sep 25
Handling licenses attached to data
- CS
- Open source, collections, open-weight model licenses
- Law
- Copyright basics, licenses, first sale doctrine
- Cases
- Vernor v. Autodesk, Jacobsen v. Katzer, MDY v. Blizzard, Doe v. GitHub, Bartz v. Anthropic (skipping the fair use order for now)
Week 4
Oct 2
Copying within the data pipeline
- CS
- Hardware/software data pipeline: memory, disk, copies
- Law
- Fixation, intermediate copying, memorization
- Cases
- Feist v. Rural, Sega v. Accolade, Cartoon Network v. CSC Holdings, MAI Systems v. Peak, Andersen v. Stability AI
Week 5
Oct 9
Training the model and fair use
- CS
- Representation: tokenization and embeddings
- Law
- Fair use factors, what is transformative
- Cases
- Authors Guild v. Google, Warhol v. Goldsmith, Thomson Reuters v. Ross Intelligence, Google v. Oracle, Kadrey v. Meta Platforms, revisit Bartz for fair use
Week 6
Oct 16
Violations in the model output
- CS
- Style transfer, deepfakes, abstraction-filtration-comparison
- Law
- Substantial similarity, trademarks, right of publicity, derivative works
- Cases
- Jack Daniel's v. VIP Products, Midler v. Ford, Lehrman v. Lovo, New York Times Co. v. OpenAI, Disney v. Midjourney, Cox v. Sony
Week 7
Oct 23
Privacy and active training
- CS
- Post-training, RLHF, distillation, and human review
- Law
- Privacy law, trade secrets, and confidentiality
- Cases
- ACLU v. Clearview AI, USA v. Heppner, re Google Assistant Privacy Litigation, xAI Corp v. OpenAI, Brewer v. Otter.ai
Week 8
Oct 30
Search and attribution
- CS
- Retrieval-augmented generation and web search
- Law
- Display and attribution, substitution, agency law
- Cases
- Perfect 10 v. Amazon, Dow Jones v. Perplexity AI, Associated Press v. Meltwater, Amazon v. Perplexity AI, Watteau v. Fenwick
Week 9
Nov 6
Authoring, and what it means
- CS
- Prompting, predictability
- Law
- Creativity, human vs. AI authorship, inventorship
- Cases
- Burrow-Giles v. Sarony, Thaler v. Perlmutter, Zarya of the Dawn, Allen v. Perlmutter, Thaler v. Vidal
Week 10
Nov 13
Responsibility for generated content
- CS
- Hallucinations, context, drift
- Law
- Section 230, defamation, free expression
- Cases
- Walters v. OpenAI, Zeran v. America Online, Mata v. Avianca, Moody v. NetChoice, Moffatt v. Air Canada, New York Times v. Sullivan
Week 11
Nov 20
Torts for serious harms
- CS
- Alignment, guardrails, jailbreaking, and chatbots
- Law
- Product or service? Tort framework: negligence, duty, causation, damages; product liability law
- Cases
- Raine v. OpenAI, Garcia v. Character Technologies, Lemmon v. Snap, Greenman v. Yuba Power Products, James v. Meow Media
Week 12
Thanksgiving
Week 13
Dec 4
Regulations
- CS
- Watermarking and auditing
- Law
- State and EU statutes/regulations
- Cases
- Mobley v. Workday, EU AI Act, Colorado AI Act, California statutes, OpenAI v. Garante
Learning Activities and Assessments
Leading Discussion11
Discussion Notes 10 submissions · 2 points each20
Discussion Reflections 6 submissions · 2 points each12
Assignment30
LexFactor 9 sessions · 3 points each27
Letter grades are calculated automatically at the end of the semester. The thresholds for A/B/C cutoffs are 90/80/70. Historically, the distribution of grades is roughly in line with the distribution in our department or division; see the Brown Courses Factbook for the breakdown. Neither the course nor the assignments are graded on a predetermined distribution; nothing is graded on a curve.
Leading Discussion
Students will choose a week in which they will lead the discussion during class. Prepare notes for how the discussion should go, including a very short summary of the readings, open-ended prompts for the class, timely information that is interesting to share, and a plan for the flow of the discussion.
Based on the topic for that week, also introduce a hypothetical scenario with a legal question where you expect to stump the class: try to get half the class to think one way and the other half to think the other way.
Discussion Reflections
Students will submit reflections for their discussion contributions in six classes, no more than 500 words each. Describe what was being discussed before the contribution, what you contributed and why, and how it related to the class topic or reading. Reflections should be written by the student without the assistance of AI tools.
Assignment
An assignment will be released during the middle of the semester and will be due before Thanksgiving break. It will ask students to provide a legal analysis of a technical scenario reported in current events.
Discussion Notes
Each week, prepare a brief comprising handwritten notes on paper that outline the key issues of the cases before class. Include the legal question, key facts, procedural history, relevant law and its application, the court's holding, and your thoughts or questions.
Notes do not need to be clean or long. You may earn up to 2 points for participation in each class by submitting a clear photograph or scan of your notes to Canvas before class. Notes should reflect your own ideas; do not submit AI-generated notes. Students with an approved accommodation that makes paper-based participation difficult should speak with the instructor to identify an alternative.
LexFactor
Readings will be available in LexFactor with the relevant parts excerpted. LexFactor sessions are conversations with an AI “professor” about the reading, guided by questions that address the key issues of the cases. Nine sessions are included in the grade.
LexFactor work is open book and has no time limit, but the use of AI tools is not permitted. Assessment is based on the student's own contributions during the conversation. Like a graded reading comment, credit reflects what the student personally writes about the key issues in the reading, while the exercise also functions as an assessment like a quiz with guided questions. The complete transcript of each session is shared with the course staff, who finalize a grade based on the key issues covered by the student. Each session is expected to take approximately 1–2 hours; complete the assigned reading before beginning. Access LexFactor at lexfactor.cs.brown.edu.
Frequently Asked Questions
Can I use generative AI in this course?
In the context of this course, “generative AI tools” include agents such as ChatGPT, Claude, Gemini, Copilot, Perplexity, Mistral, and other systems that generate or transform text, ideas, code, images, and other content.
Unless an assignment or activity explicitly states otherwise, you may not use generative AI tools in any capacity to complete course assignments, including during discussions and on LexFactor.
As this course focuses on the legal implications of generative AI use, it is important for you to develop your own abilities to understand, reason, research, and construct your own analyses and arguments. Law and computer science are detail-oriented fields that emphasize one's ability to think critically and question existing assumptions. The assignments are designed to assess your legal reasoning, not your writing or coding abilities. Having an AI system brainstorm or analyze court cases for you would undermine your ability to develop these skills.
Understanding when and how to use AI appropriately is an important aspect of this course. However, submitting AI-generated or AI-assisted work as your own may constitute a violation of the University's academic integrity policies. The prohibition applies even if you have substantially edited, rewritten, or cited AI-generated material, as the goal is to ensure that submitted work reflects your own thought processes and ideas. If you have questions about what tools are allowed, ask the instructor as soon as the assignment has been introduced.
What happens if I miss class?
A substantial portion of learning takes place during class meetings and discussion. Students who miss more than one week should expect their grade to be affected, as there are limited opportunities to make up missed work.
Can I bring my laptop, tablet, or phone to class?
The course uses a largely paper-based notetaking and discussion format. Laptops, tablets, and phones should be put away except for an approved accommodation or a specific class activity. Bring your prepared notes to class on paper.
What if I submit something late?
Discussion reflections, assignments, and LexFactor sessions lose one course point for each day late. A LexFactor session one day late can receive at most 2 of 3 points; an assignment four days late can receive at most 26 points.
How do I enroll?
An override request form will be made available on the course website following the first class meeting on Friday and must be submitted by the end of the following Monday. Override codes will be distributed by the following Wednesday evening to 20 students, before the second class on Friday. No additional override codes will be distributed after Wednesday.
Students do not need to submit an override request through CAB at this time or contact the instructor by email. There are no prerequisites; any student enrolled at Brown may request to enroll and will be considered.
How much work is this course?
The course is expected to take 180 hours of work during the semester, comprising weekly class, discussion reflections, readings, an assignment, and LexFactor sessions.
Will there be sensitive content?
Due to the nature of the course, students will encounter content about harm and other sensitive topics. Some court cases may describe suicide, abuse, or death in classes about chatbot liability and harm.
How do I request accommodations or accessibility support?
Brown University is committed to full inclusion of all students. Please inform me early in the term if you may require accommodations or modification of any course procedures. You may speak with me after class, during office hours, or by appointment.
Student Accessibility Services (SAS) provides accommodations and support for undergraduate and graduate students with disabilities. To learn more or request disability accommodations, email sas@brown.edu, call 401-863-9588, or visit the SAS website.
If you encounter inaccessible digital materials in this course, complete the Digital Accessibility Concern Reporting Form. For other accessibility or accommodation concerns, contact the ADA/504 Coordinator at ada_504@brown.edu.
Undergraduates in need of short-term academic advice or support can contact an academic dean in the College at college@brown.edu. Graduate students may contact a dean in the Graduate School at graduate_school@brown.edu.
Can I take an Incomplete?
A grade of Incomplete will be considered only when a single requirement remains outstanding at the end of the semester, and only with a Dean's Note specifically recommending an Incomplete for good cause, together with the approval of the instructor.