Is Your Website Being Used for AI? Can You Stop it?

Published Sept. 2023

Eric Robinson
By Eric P. Robinson, USC School of Journalism and Mass Communications

Chatbots such as ChatGPT and other artificial intelligence services have exploded online since their introduction just a few months ago. They have become so popular in part because of their uncanny ability to mimic human writing and thought on any topic, even if the information they provide is often wrong.

The information—and disinformation—that these services offer are largely the result of comprehensive web surfing and collection of material from online sources. Since most of the material online is protected by copyright, this has led many to question what websites can and should do about this, including asking whether use of this material as fodder for chatbot and AI programs is somehow copyright infringement.

In recent months, several copyright infringement lawsuits have been filed against AI companies’ use of material created by others, including legal claims filed by authors, computer code writers, stock photo agencies and visual artists.

Could newspaper publishers join them? (The New York Times, for one, has reportedly considered filing its own lawsuit.)

There are four ways that AIs and chatbots use material found online: collecting the material from various websites; retaining the material; analyzing the material and using it to “train” their language skills; and using the material as a collection of “facts” for the production of text and/or graphical responses to user queries and instructions.

Let’s address the legalities of each of these:

Whether the collection of copyrighted material by chatbots and AIs is infringement is an open legal question. In non-digital activities, after all, “mere” collection of research data is not in and of itself infringement, unless the collection involves making an unauthorized copy of the source data. So, for example, a researcher can use facts from a newspaper article in her research, and use that information by quoting portions of the article or paraphrasing it in the final product in a way that does not substitute for the original article. (Of course, ethics would likely require citation of the article as a source.) These would likely be considered “fair use.” But she could not just copy the text of the article and present it as her own; that would be copyright infringement.

In the context of AIs and chatbots, the collection of web content as source material itself would likely not be copyright infringement. There could be a legal claim if such materials were collected in violation of websites’ terms of service, such as terms of service provisions barring unauthorized “scraping” of the sites or barring use of information on the site for commercial purposes, but the terms of service would actually have to explicitly include such provisions; and this claim would be based on contract law—violating the terms of service—not copyright.

Retention of material found online may indeed be a copyright infringement, in the same way that making and circulating unauthorized copies of articles for informational purposes within a firm was held to be infringement. But in that case, the Second Circuit Court of Appeals held that “… if Texaco stopped making ‘free’ photocopies, it would fill this gap through some combination of the methods discussed and, in doing so, would add significant value to the publishers’ copyrights.”  In other words, the unauthorized copying affected the publisher’s profits. In the situation of AIs and chatbots, the retention of material found online, while likely unauthorized, would likely not affect the market of the source materials: at least, until it is provided to users of the AIs and chatbots in the programs’ responses, which is discussed below. Again, violation of the source websites’ terms of service would be another possible, non-copyright, claim.

Analyzing the material found online and using it to “train” AIs’ and chatbots’ language skills presents a new challenge under copyright law. For example, use of copyrighted materials for analysis—with appropriate citation—is common in areas such as academia, law and journalism, and may be considered “fair use” under copyright law. But is use of these materials by AIs and chatbots the same as use by journalists and others who perform functions we and the law deem to be socially valuable?

It is regarding the final issue—inclusion of copyrighted material in chatbot and AI output without owners’ permission—where established copyright law provides the clearest direction. The copyright statute provides guidance on whether such reuse can be considered “fair use,” or whether a new work incorporating pre-existing works is so different from the pre-existing works that it is “transformative” (technically, a type of “fair use”). For example, in 2015 the Second Circuit held that the Google Books project, which digitized hundreds of thousands of books held by participating libraries and made them word-searchable, was transformative. On the other hand, the digitization of archives of The New York Times and other newspapers was held not to be transformative. The question is whether AI and chatbot output that includes copyrighted materials culled online will fit into either the “fair use” or “transformative” categories.

The lawsuits filed against AI and chatbot companies may answer these questions. Or Congress may step in. But it is clear that these new technologies present clear challenges to existing concepts of copyright, and that lawmakers and the courts will likely be struggling with these changes for many years to come.

In the meantime, online publishers should do what they can to ensure that their terms of service explicitly cover these activities. Of course, such changes will not automatically stop misuse of online material. But adding such provisions will give online publishers a legal basis for taking action when it occurs.

A Resounding Endorsement of Open Courts

On Sept. 6 the South Carolina Supreme Court released its opinion explaining its brief order issued more than four months earlier vacating the early release of convicted murderer Jeroid John Price after serving 19 years of a 35-year sentence, and ordering him back to prison. (Price was re-arrested in July.)

Price benefited from a South Carolina law that allows early release for prisoners who assist prison authorities. His attorney, J. Todd Rutherford—who is also minority leader of the South Carolina House of Representatives—apparently made a motion for the early release in February 2022, and it was granted by now-retired Circuit Court Judge Casey Manning on Dec. 30, 2022; his last day in office. The entire process was sealed, with no public record of the motion, of any hearings held on the matter, or of the order granting release. Price was released on March 15, 2023. Press reports about the release were published a month later, leading the Supreme Court to order on April 20 that all records relating to the case be unsealed and on April 26 to issue its order revoking the release.

In the Sept. 6 decision explaining the April 26 order, the three-judge majority held that then-Justice Manning lacked authority to issue the sentence reduction order because the process was done in secret. “By signing the ‘Order Reducing Sentence’ without … compliance with the … requirement of filing a written motion seeking the reduction of the inmate’s sentence, without a copy of the motion having been provided to the chief judge of the circuit, and without otherwise complying with the statutory [disclosure] requirements,” the majority opinion by Associate Justice John C. Few states, “Judge Manning committed multiple errors of law and acted outside his authority.” The majority also noted that Judge Manning had not sought any alternatives to closure, as required by U.S. Supreme Court and South Carolina Supreme Court precedent. The opinion also cites the general requirements for open courts in the U.S. Constitution and the South Carolina Constitution and statutes.

The Court’s majority opinion is imbued with statements on the importance of public access to the courts, including quoting a U.S. Supreme Court statements that “[t]he knowledge that  every  criminal [proceeding] is subject to contemporaneous review in the forum of public opinion is an effective restraint on possible abuse of judicial power,” and quoting a 19th century British legal scholar stating, “[S]uppose the proceedings to be  completely secret, and the court, on the occasion, to consist of no more than a single judge—that judge will be at once indolent and arbitrary: how corrupt soever his inclination may be, it will find no check, at any rate no tolerably efficient check, to oppose it. Without publicity, all other checks are insufficient: in comparison of publicity, all other checks are of small account.”

While the dissenting opinion by Justice George C. James, Jr. argued that the Supreme Court “should not grant the State relief from the sentence-reduction order just because the State regrets the result it sought and obtained in the circuit court,” it agreed with the majority on the importance of open courts. “There is certainly unanimity in this Court that any court should exercise great care and discretion before closing proceedings or sealing any portion of a court record,” James wrote. But he added, “In such cases, the appropriate remedy is to vacate both the order closing the proceedings and the order sealing documents, not to undo the proceedings” (original emphasis). “This Court should not relieve the State from its obviously poor choice to concoct a secret plan to seek a reduction for Price.”

Alex Murdaugh Sanctioned for Media Interview

On Aug. 30, the South Carolina Department of Corrections announced that it had punished convicted murderer Alex Murdaugh for covertly conducting a media interview in violation of its policy not allowing any inmate interviews, as well as for using a fellow inmate’s phone PIN number after his telephone privileges were revoked while the interview was investigated.

The interview for the Fox Nation documentary “The Fall of the House of Murdaugh” consisted of Murdaugh reading entries from a journal he kept during his trial. Murdaugh was recorded reading the entries during a phone call with his attorney Jim Griffin, who provided the recordings to the network. While prisoners’ phone calls are routinely monitored and recorded—with the recordings subject to disclosure under the state’s Freedom of Information Act—calls with attorneys are exempt.

In a letter to Griffin, South Carolina Department of Corrections Senior Assistant Deputy Director of Operations Dennis R. Patterson wrote that “[Griffin’s] actions, whether … intended or not, assisted Mr. Murdaugh in violating our policy and could jeopardize your telephonic communications with him in the future. Attorney calls are provided to assist with legal claims, not for other unrelated purposes.”

There is no First Amendment right to inmate interviews, although the American Bar Association recommends that prison authorities accommodate such interviews. The states have a wide variety of policies, with South Carolina’s total ban on interviews being the most restrictive. “SCDC’s interview policy is rooted in victims’ rights and is longstanding,” the department said in its press release. “The department believes that victims of crime should not have to see or hear the person who victimized them or their family member on the news. Inmates lose the privilege of speaking to the news media when they enter SCDC.”

The press release also notes that while Murdaugh was punished for disregarding prison rules, his transgressions were not violations of law. Meanwhile, Murdaugh is appealing his conviction, seeking a new trial based on the argument that the court clerk improperly influenced the jury deliberations in anticipation of publishing a book on the trial.

Eric P. Robinson focuses on media and internet law as associate professor at the USC School of Journalism and Mass Communication and in an “of counsel” position at Fenno Law in Charleston / Mount Pleasant. He has worked in media law for more than 25 years and is admitted to legal practice in New York and New Jersey and before the U.S. Supreme Court. This column is for educational purposes only; it does not constitute legal advice. Any opinions are his own, not necessarily those of his employers.

Other recent columns