Inside AI Training: Copyright Liability from Shadow Libraries to RAG Outputs

Miranda Means, Partner | Kirkland & Ellis LLP

Miranda Means is a partner in the Boston office of Kirkland & Ellis LLP, where her practice spans litigation and counseling across copyright, trade secret, trademark, right of publicity, art, internet, and advertising law. She represents clients in disputes touching a remarkable breadth of media—social media, video games, fine art, sports, music, and film—and the technologies reshaping them, from artificial intelligence and computer software to consumer electronics. Miranda is perhaps best known for her role in the first AI copyright fair use case in U.S. legal history, work for which American Lawyer named her a “Litigator of the Week.” She pairs a sharp litigation record with a deep commitment to inclusivity and pro bono service, including matters advancing LGBTQ+ rights.

• Education & Credentials

Miranda earned her J.D. magna cum laude from Harvard Law School in 2017, where she was a Dean’s Scholar in Copyright; Music and Digital Media; Legal Writing and Research; and the Fashion Law Lab. She served as Articles Editor and on the Submissions Committee of the Harvard Journal of Sports and Entertainment Law, was a Teaching Fellow for Professor William Fisher’s CopyrightX, and worked as a student attorney with both the Prison Legal Assistance Project and the Cyberlaw Clinic at the Berkman Center for Internet & Society. She received her B.A. in English Language and Literature with Honors from the University of Chicago in 2014, where she was elected to Phi Beta Kappa and named to the Dean’s List every quarter. She is admitted to practice in Massachusetts (2020) and New York (2018) and before the U.S. Courts of Appeals for the Second, Third, Sixth, and Eleventh Circuits, as well as the U.S. District Courts for the District of Massachusetts and the Southern District of New York.

• Recognition & Leadership

Miranda was recognized by American Lawyer as a “Litigator of the Week” for her work on Thomson Reuters v. ROSS Intelligence, the first decision in U.S. legal history addressing the use of copyrighted material to train an AI model. She was named a Super Lawyers “Rising Star” each year from 2020 through 2024. Within the American Bar Association’s Section of Intellectual Property, she served as Vice Chair of the Visual Arts and Dramatic Works Committee from 2023 to 2024.

• Professional Involvement

Miranda is an active member of the American Bar Association Section of Intellectual Property and the International Trademark Association (INTA), where she served on the Internet Committee’s Internal Communications and Research Subcommittee. She is a frequent speaker on intellectual property and emerging technology, with recent engagements including a panel on “Lawyering and AI” at Columbia Law School, an IP Law Update panel at Lavender Law’s IP Law Institute, and guest lectures on tattoo copyright and video game IP at New England Law, Harvard Law School, and UCLA Law. Her published work includes articles in the New York Law Journal and the ABA’s Landslide. She is also dedicated to pro bono work, including matters protecting LGBTQ+ rights.

• Experience

Miranda has built a litigation record across many of the most closely watched IP disputes at the intersection of media and technology. She won summary judgment for Thomson Reuters and West Publishing in Thomson Reuters v. ROSS Intelligence, the landmark ruling holding that copying copyrighted material to train a competing AI model was not fair use. She secured a complete victory for Astronics Test Systems in a multi-front patent, copyright, and state-law dispute against Teradyne—prevailing on motions to dismiss, invalidating a patent through inter partes review, and winning a fair use ruling later affirmed on appeal. In the video game arena, she won a jury trial for 2K Games in Hayden v. 2K Games concerning the depiction of NBA players’ tattoos and successfully defended Take-Two in related tattoo-copyright and trademark matters, including disputes over WWE 2K, Red Dead Redemption 2, and Grand Theft Auto: Vice City.

Her broader practice includes defending Facebook in trademark litigation, representing a leading consumer electronics company through a Second Circuit affirmance, and handling trade secret and Computer Fraud and Abuse Act claims for clients such as Apollo Aviation Group and Dow Chemical. On the transactional and counseling side, she represents Vista Equity Partners in worldwide trademark prosecution and enforcement, conducts trademark clearance across industries ranging from pharmaceuticals to professional athletics, and advises clients on artificial intelligence policy and enforcement.

Zahr K. Said, Professor of Law | Santa Clara University

Zahr K. Said is a Professor of Law at Santa Clara University School of Law, where her scholarship sits at the intersection of copyright, artificial intelligence, and the interpretive questions at the heart of intellectual property disputes. A nationally recognized voice on the role of the jury in IP litigation and on how courts read and construe creative works, Professor Said brings an unusually interdisciplinary lens to legal questions—grounded in advanced training in comparative literature as well as law. She is also a committed teacher and innovator in legal pedagogy, having authored an open-access tort law casebook that reframes the field around questions of race, gender, class, and ability.

• Education & Credentials

Professor Said earned her J.D. from Columbia University Law School. She holds a Ph.D. in Comparative Literature from Harvard University and a B.A. in Comparative Literature from the University of California, Berkeley. Her areas of specialization include copyright law, tort law, law and the humanities, and advertising law.

• Recognition & Leadership

During her tenure at the University of Washington School of Law, Professor Said received the 2015 Law Faculty Scholarship Award and was named the 2016 Philip A. Trautman 1L Professor of the Year, alongside a grant from the University’s Global Innovation Fund. She currently serves on the editorial board of the journal of the Copyright Society of America, reflecting her standing in the copyright field. Her open-access tort law casebook—now in its second edition—has been recognized as a notable contribution to inclusive and accessible legal education.

• Professional Involvement

Professor Said is an editorial board member of the Copyright Society of America’s journal and a frequent participant in scholarly and professional convenings on copyright and emerging technology, including conferences marking the legacy and future of U.S. copyright law. Through her open-access publishing and her work on AI in legal education, she is actively engaged in broadening access to legal scholarship and shaping conversations about how the profession adapts to new technologies.

• Experience

Professor Said joined the Santa Clara Law faculty in 2024, following thirteen years on the faculty of the University of Washington School of Law. She has also taught at the University of Virginia School of Law and Stanford Law School. Her teaching centers on tort law, and her research program spans copyright law, artificial intelligence, remedies, and the interpretive challenges that arise when courts evaluate creative and expressive works—with particular attention to the jury’s role in intellectual property litigation. She is currently writing a book on artificial intelligence and legal education, and her published open-access tort casebook reorients the teaching of the subject around race, gender, class, ability, and other sociological dimensions.

SESSION 1 – AI Training & Copyright: Direct and Contributory Liability | 1:00pm – 2:00pm

This session traces the LLM training pipeline—from shadow-library acquisition through tokenization, pretraining, fine-tuning, and RLHF—mapping each stage to discrete copyright exposure under §§ 106(1) and 106(3) and analyzing Bartz, Kadrey, Andersen, and the Supreme Court’s Cox decision.

BREAK | 2:00pm – 2:10pm

SESSION 2 – Copyright Risks After AI Training: RAG and Outputs | 2:10pm – 3:10pm

Shifting past training, this session examines copyright risks in runtime retrieval and output generation—substantial similarity, RAG architectures, DMCA § 1202 CMI claims, the § 512 safe harbor, and litigation including Getty, Disney v. Midjourney, and NYT v. OpenAI.