A recent opinion dismissing a lawsuit over AI training gets it right on standing: harmless acts don’t belong in federal court. Last week, Judge Colleen McMahon in the Southern District of New York ruled that publishers Raw Story and Alternet lacked Article III standing to sue OpenAI because they had not alleged “concrete harm,” a constitutional requirement interpreted most recently in the Supreme Court’s TransUnion v. Ramirez opinion. The publishers said OpenAI had violated the statute by removing “copyright management information” (or CMI) from their stories during the AI training process, but Judge McMahon said the publishers didn’t accuse OpenAI of causing them any concrete harm, so there was no “case or controversy” for the court to decide. In short, if OpenAI’s behind-the-scenes training activities didn’t result in any publication of infringing material, where’s the harm? You can read more about the Raw Story opinion on Aaron Moss’s excellent Copyright Lately blog. I want to focus on why the TransUnion “concrete harm” test is good copyright policy.

The law forbids removal of CMI in situations where the user knows that doing so will enable copyright infringement—a paradigm case might be someone who copies an entire book, strips the CMI to cover their tracks, and posts the book for sale online, where every sale would be an infringement. By contrast, in the Raw Story case Judge McMahon pointed out that there was little reason to believe that OpenAI’s removal of CMI would result in similar concrete harm to Raw Story or Alternet. Because ChatGPT is trained on millions of works, it is unlikely to reproduce any particular story verbatim (without CMI) in response to a user’s query. This reasoning is quite similar to Judge Leval’s fair use reasoning in Authors Guild v. Google: there may be a lot of copying going on behind the scenes, but the public result is a new and useful information tool that doesn’t provide infringing copies to the public. No harm, no foul. 

“No harm, no foul” is an apt approach to standing for copyright law. Under the Constitution, copyright is a means to an end: exclusive rights are granted for limited times “to promote the Progress of Science and the useful Arts.” Copyright’s limits on public access and use of in-copyright works are necessary evils, only justified to the extent that they provide an incentive for creation of new works, which in turn benefit the public. Uses that benefit the public without harming the creator should be (and often are) encouraged by the law, not discouraged. Fair use, the public domain, rights for libraries and archives, and DMCA exemptions are all protections for uses that benefit the public and do no harm to the copyright holder. 

Of course some copyright holders want to squeeze every last drop of profit from their rights (just ask Miley, Ed, Taylor, or these folks), but courts are under no obligation to treat every lost licensing opportunity as a “concrete harm.” In Oracle v. Google, Justice Breyer explained that fair use “provid[es] a context-based check that can help to keep a copyright monopoly within its lawful bounds.” The TransUnion rule on Article III standing serves the same important role by kicking harmless copyright cases out of court.