With the emergence of large language models (LLMs) like ChatGPT, scholars and courts have fervently debated whether LLMs’ training on and reproduction of copyrighted materials amounts to fair use. But in a recent series of cases, a lesser-known challenge to LLMs has reared its head: § 1202 of the Digital Millennium Copyright Act. This provision requires that when a work is copied, its associated copyright management information (CMI)—such as its license or terms of use agreement—be copied with it. This requirement was originally intended to modernize copyright for the internet by ensuring that all users would be aware of the terms of their use. Now § 1202’s unintended overbreadth threatens to block LLM development and use as it swallows questions of infringement and fair use entirely.

This Note posits that § 1202 is broader than traditional copyright infringement doctrine in three critical respects: It imposes liability without any showing of copyrightability, provides no fair use defense, and permits disproportionate statutory damages. Although § 1202 includes an intent requirement, courts have applied it so minimally that it fails to constrain the statute’s reach—especially in the LLM context, where the mere act of violating § 1202 may itself suffice to establish intent.

To restore meaningful limits to § 1202, this Note proposes that courts adopt an identicality requirement for § 1202 claims against LLMs. The requirement would cabin liability to outputs that exactly match training data—cases where the removal or alteration of CMI is both clear and technically avoidable. This approach mirrors Congress’s existing accommodation of industry limitations for broadcasters in cases of technical infeasibility or financial hardship. A similar understanding for LLMs would preserve § 1202’s core purpose, resolve a growing district court split, and ensure that the most consequential copyright question raised by generative artificial intelligence is answered on its merits—not sidelined by a statute never meant to decide it.

* Stanford Law School, J.D. 2025. Thank you to Phil Malone, Nina Srejovic, Braden Crimmins, and the Juelsgaard Intellectual Property and Innovation Clinic for the teamwork, inspiration, and ideation that led to this Note. Thank you also to the outstanding editors of the Stanford Law Review. All errors are my own.