Is Data Scraping for AI Considered “Fair Use”?

Attorney Nick Oberheiden
AI Litigation Team Lead

Artificial intelligence (AI) developers need data to train their AI platforms. The more data they can feed into their platforms, the “smarter” their platforms will become. This, in turn, will lead to better, more accurate, and more consistent outcomes; and, ultimately, it will give their AI platforms the foundation needed to innovate and generate new data on their own.

But, is this a “fair use” of third parties’ publicly available data?

“Fair use” is a copyright doctrine that allows parties to use other parties’ copyrighted works without authorization in certain circumstances. It creates an exception to the general rule that a license is required to use someone else’s (or a company’s) copyrighted works. When “fair use” applies, not only is the user not required to obtain a license, but the user is not required to inform the copyright owner that its protected works are being used.

Copyrighted Works vs. Unprotected Ideas and Information

Before we get into the “fair use” implications (or lack thereof) of data scraping for AI training purposes, we first need to highlight an important distinction that often leads to confusion. This is the distinction between copyrighted works and unprotected ideas and information.

In the United States, federal copyright law protects “original works of authorship fixed in a tangible medium of expression.” While original works of authorship necessarily reflect new ideas or contain information, these ideas and information are not the focus of the U.S. Copyright Act. Rather, it is the creative expression of ideas or information that is protected—and a fixed medium is necessary to have something that is both eligible for protection and capable of being protected. As Princeton University explains:

“The Copyright Act . . . protects all kinds of expression or authorship that are fixed in tangible medium – including written works, paintings, sculptures, photographs, videos, recorded music, sheet music, computer programs, video games, architectural design, and choreography. The Copyright Act DOES NOT protect facts, data, or ideas. But it does protect the fixed expression of ideas.

“Copyright protection starts when the expression is fixed in a tangible medium, regardless of whether the work has been published, registered with the U.S. Copyright Office, or marked with a copyright notice.”

Taking this into account, it might initially seem as though data scraping for AI training purposes doesn’t implicate the U.S. Copyright Act at all. AI platforms feed on data—the information and ideas that are publicly available on the internet—rather than feeding on how these data are presented.

But, in order to process publicly available data, AI platforms need to get the data from somewhere. This is where federal copyright law comes into play. When data scraping, AI platforms rely on third parties’ original works of authorship (i.e., websites, videos, and other online content), and they create “derivative works” that bring copyright owners’ exclusive rights into play.

Dr. Nick Oberheiden

Founder

Attorney-at-Law

Lynette S. Byrd

Former DOJ Trial Attorney

Partner

Brian J. Kuester

Former U.S. Attorney

Hon. Kevin McCarthy

55th Speaker, U.S. House of Representatives (ret.)

Government Consultant

Mike Pompeo

Of Counsel

Former U.S. Secretary of State

John W. Sellers

Former Senior DOJ Trial Attorney

Linda Julin McNamara

Federal Appeals Attorney

Nicholas B. Johnson

Former Prosecutor

Roger Bach

Former Special Agent (DOJ)

Chris J. Quick

Former Special Agent (FBI & IRS-CI)

Michael S. Koslow

Former Supervisory Special Agent (DOD-OIG)

Ray Yuen

Former Supervisory Special Agent (FBI)

The Doctrine of “Fair Use”

In light of the copyright implications of data scraping for AI training purposes, AI developers essentially have two options when it comes to using data in third parties’ content: (i) obtain a license; or, (ii) rely on the doctrine of “fair use.” So, what is “fair use”?

As we mentioned above, “fair use” is a doctrine that creates an exception to copyright owners’ exclusive rights. As the U.S. Copyright Office explains:

“Fair use is a legal doctrine that promotes freedom of expression by permitting the unlicensed use of copyright-protected works in certain circumstances. Section 107 of the Copyright Act provides the statutory framework for determining whether something is a fair use and identifies certain types of uses—such as criticism, comment, news reporting, teaching, scholarship, and research—as examples of activities that may qualify as fair use. Section 107 calls for consideration of . . . four factors in evaluating a question of fair use.”

The four factors used to determine whether an unlicensed use of a copyrighted work constitutes “fair use” are:

1. The Purpose and Character of the Use

The first factor is the purpose and character of the use in question. In the words of the U.S. Copyright Office, “Courts look at how the party claiming fair use is using the copyrighted work, and are more likely to find that nonprofit educational and noncommercial uses are fair.” However, as the Copyright Office also makes clear, “[t]his does not mean, however, that . . . all commercial uses are not fair.” As the Copyright Office also notes, “transformative” uses that, “add something new, with a further purpose or different character, and do not substitute for the original use of the work,” are “more likely to be considered fair.”

2. The Nature of the Copyrighted Work

The second factor is the nature of the copyrighted work. As the Copyright Office explains, this factor “analyzes the degree to which the work that was used relates to copyright’s purpose of encouraging creative expression. . . . [and] using a more creative or imaginative work (such as a novel, movie, or song) is less likely to support a claim of a fair use than using a factual work (such as a technical article or news item).”

3. The Amount and Substantiality of the Portion Used

The third factor is the amount and substantiality of the portion of the copyrighted work that is being used. As the Copyright Office explains, “courts look at both the quantity and quality of the copyrighted material that was used.” While this generally means that using a substantial portion of a copyrighted work (or copying a protected work wholesale) is more likely to be deemed infringing, “using even a small amount of a copyrighted work [may be] determined not to be fair [if] the selection [is] an important part—or the ‘heart’—of the [protected] work.”

4. The Effect on the Potential Market or Value of the Copyrighted Work

The fourth factor is the effect on the potential market or value of the copyrighted work. “Here, courts review whether, and to what extent, the unlicensed use harms the existing or future market for the copyright owner’s original work.” This factor can weigh against a finding of fair use if either: (i) the unlicensed use “is hurting the current market for the original work;” or, (ii) “could cause substantial harm if it were to become widespread.”

It is also worth mentioning that copyright protection expires—and some once-copyrighted materials are no longer protected. Generally, copyright protection lasts for the life of the author plus 70 years. For example, Disney’s copyright for the original depiction of Mickey Mouse expired in early 2024. However, future depictions of Mickey Mouse remain protected, and this presents challenges for parties seeking to claim non-infringing use—not only with regard to Mickey Mouse, but with regard to expired copyrights generally.

The Argument For “Fair Use” in AI Data Scraping

With this in mind, what is the argument for fair use in AI data scraping? This argument is primarily based on the first of the four factors listed above—with particular emphasis on the “transformative” nature of the use. However, to date, even many of the major AI developers have struggled to clearly articulate a basis for their claim. For example, in response to a copyright infringement lawsuit filed by The New York Times, OpenAI responded that, “training AI models using publicly available internet materials is fair use, as supported by long-standing and widely accepted precedents,” and, “based on well-established precedent, the ingestion of copyrighted works to create large language models or other AI training databases generally is a fair use.”

These general claims do little to advance the company’s argument; and, since facing The New York Times’ lawsuit, OpenAI has reportedly sought to license content from various entities. However, the case for fair use is far from over, and it remains to be seen how courts will resolve the issue across the country.

The Argument Against “Fair Use” in AI Data Scraping

Now, what about the argument against fair use? Arguably, all four of the “fair use” factors weigh against unlicensed data scraping for AI training purposes. The use is commercial in nature, it covers all types of protected works, it involves the copying of entire protected works, and it eliminates an entire potential market for copyright owners. But, the law is complex and individual circumstances are unique; and, as a result, developers are likely to assert a wide range of “fair use” claims that will ultimately need to be decided in AI litigation on a case-by-case basis.

If your company develops AI using third-party data, or if you are a copyright owner affected by AI data scraping, what does all of this mean for you? The short answer is that you need to ensure that you are making informed decisions based on the guidance that is currently available. If you have questions, we can help, and we invite you to contact us for more information.

Schedule a Complimentary Consultation at Oberheiden P.C.

Our attorneys have extensive experience advising clients on all types of AI-related matters. To schedule a complimentary consultation with a senior attorney at Oberheiden P.C., please call 888-680-1745 or request an appointment online today.