Authors Guild, Inc. v. HathiTrust, No. 12‐4547 (2d Cir. June 10, 2014)
Several research universities allowed Google to scan their collections; then they created a repository for the digital copies, the HathiTrust Digital Library (HDL), which currently has 80 member institutions and over ten million works. HathiTrust allows the public to search for particular terms. Unless otherwise authorized, the search results show only page numbers and the number of times the term appears, with no snippets, like so:
HathiTrust also allows member libraries to provide patrons with certified print disabilities access to the full text of copyrighted works. Print disabilities include blindness and disabilities that prevent a person from physically holding a book or turning pages. Print‐disabled users can obtain access using adaptive technologies such as software that converts the text into spoken words, or that magnifies the text. Michigan was the only member that permitted such access; others intended to provide it in the future, which, presumably, starts now.
Finally, by preserving the copyrighted books in digital form, the HDL plans to permit members to create a replacement copy of the work, if the member already owned an original copy, the member’s original copy is lost, destroyed, or stolen, and a replacement copy is unobtainable at a “fair” price elsewhere.
Michigan also developed an Orphan Works Project that would attempt to identify out of print works, try to identify their copyright owners, and, if none could be found, publish a list of orphan works candidates. If no owner still came forward, orphan works would be made accessible in digital format to library patrons, with simultaneous viewers limited to the number of hard copies owned by the library. Michigan became concerned that its screening process wasn’t adequately distinguishing between orphans and non-orphans, and the project was indefinitely suspended.
The district court found that the three HDL uses were fair, and that the provision for print-disabled patrons was permitted by the Chafee Amendment, which allows “authorized entities” to reproduce or distribute copies of a previously published, nondramatic literary work in specialized formats exclusively for use by the blind or other persons with disabilities. The district court concluded libraries of educational institutions have “a primary mission” to provide specialized services to print-disabled individuals, as required for eligibility to take advantage of the Amendment. The Authors Guild, for reasons that apparently seemed sufficient, appealed.
Initially, the court of appeals found that three authors’ associations (the Authors Guild, Australian Society of Authors Limited, and Writers’ Union of Canada) lacked standing to bring claims on behalf of their members. Four foreign associations, however, asserted that foreign law conferred on them exclusive rights to enforce their foreign members’ copyrights, and the libraries didn’t contest that claim. (Though I’ve got to wonder about that “exclusive.”) Thus, they had standing.
Fair use: Plaintiffs argued that §107 couldn’t apply because §108 governs reproduction by libraries. But §108 says, “Nothing in this section in any way affects the right of fair use as provided by section 107,” so that’s that.
The court went through a number of fair uses: book reviews with quotes “to illustrate a point and substantiate criticisms”; biographers quoting from unpublished works ditto; art “employ[ing]” photos “in a new work that uses a fundamentally different artistic approach, aesthetic, and character from the original”; low-resolution versions of images in search engines to direct users to the source website; newspaper use of a modeling photo to “inform and entertain” the readers for a news story; time-shifting of broadcast TV; and reverse engineering of software.
However, “[a] fair use must not excessively damage the market for the original by providing the public with a substitute for that original work.” Thus, quoting extensively from the heart of a forthcoming memoir “in a manner that usurps the right of first publication and serves as a substitute for purchasing the memoir” is not fair use.
Transformativeness is an important focus of the first factor. “A use is transformative if it does something more than repackage or republish the original copyrighted work.” However, “a use does not become transformative by making an ‘invaluable contribution to the progress of science and cultivation of the arts.’ Added value or utility is not the test: a transformative work is one that serves a new and different function from the original work and is not a substitute for it.” Plus, the impact of the use on the traditional market for the copyrighted work is the “single most important element of fair use.” “To defeat a claim of fair use, the copyright holder must point to market harm that results because the secondary use serves as a substitute for the original work.”
Full-text search: the process of full-text search requires digital copies, but doesn’t show them to users. The creation of a full-text searchable database is “a quintessentially transformative use.” The result of a word search “is different in purpose, character, expression, meaning, and message from the page (and the book) from which it is drawn.” There was “little or no resemblance between the original text and the results” of a search.
Comment: Exactly because of that last point, I don’t see how the search results could be deemed infringing at all even if there were no such thing as a fair use doctrine. Where’s the substantial similarity? Note also that the court’s focus is now on the third-party use enabled by the database, although in this particular instance the label “transformative” applies better to the database itself, since at least with the latter the reproduction right has been implicated. One way of parsing the court’s analysis would be to call the database copies intermediate copies, as in the reverse engineering cases; but the distinction is that the HDL does not perform the full-text searches, whereas reverse engineers themselves ultimately create a new work. However, perhaps because the HDL’s copying is part of a system designed so that the only output can be noninfringing, and because the HDL’s copying itself is useless unless and until that output is created, the HDL gets to take advantage of its users’ entirely noninfringing uses.
Anyhow: there was no evidence that the plaintiffs write “with the purpose of enabling text searches of their books.” Thus, full-text search doesn’t supersede the objects or purposes of the original. It wasn’t mere repackaging or republishing into a new mode of presentation. “[B]y enabling full‐text search, the HDL adds to the original something new with a different purpose and a different character.” Indeed, it adds “a great deal more to the copyrighted works at issue than did the transformative uses we approved in several other cases,” such as Cariou (same medium) or Bill Graham Archives (photos simply shrunk in size). (Interesting collapse of artforms at work here in the creation of a different kind of distinction—the court lumps together photo and collage, print and photo—but that’s a tangent.) Perfect 10 (thumbnails) and iParadigms (plagiarism detection) reinforced the court’s conclusion.
Dog that didn’t bark: no mention of commercialism. The libraries can do this with one hand tied behind their backs!
Nature of the work: it doesn’t matter, because of transformativeness.
Amount of the work: the issue is whether the copying was excessive—whether no more was taken than necessary. Here, the entire work was necessary for the purpose. Plaintiffs’ argument that the copying was excessive because the HDL maintains copies at four different locations was not well founded. They were “reasonably necessary” to facilitate legitimate uses. The two mirror sites allow for balancing user loads, and act as back-up in case of disaster. The two encrypted backup tapes were also important to protect against large-scale data loss. (The HDL also creates digital copies of images of each page; these aren’t retained for full-text search purposes and the court dealt with them under the head of print-disability access.) There was no reason to think that these copies were excessive or unreasonable in relation to the purpose.
Factor four: We’re looking for whether a use “usurps the market of the original work.” Plaintiffs could identify no harm to any “existing or potential traditional market.” The only type of harm factor four considers is substitutionary harm. “[A]ny economic ‘harm’ caused by transformative uses does not count because such uses, by definition, do not serve as substitutes for the original work.” Book reviews with quotations can deter purchases, but that doesn’t matter because they’re transformative.
Comment: it’s not hard to see how book reviews might differ from other uses that might qualify as the creation of derivative works, but a rule of this sort is necessary to avoid circularity/the collapse of fair use into a regime of universal licensing. If the collapse of fair use is something you see as a harm, then willingness to license, even innovative new or nontraditional forms of licensing, can’t be weighed against the defendant whose use is otherwise nonsubstitutionary. Also, it might be worth considering that offering a license to someone to make a new thing is not the same as having the ability to make that new thing oneself—that could be one possible boundary of “traditional markets.”
Plaintiffs offered a “lost sale” theory: a market for licensing digital search might develop in the future, but the HDL impairs its emergence. “This theory of market harm does not work under Factor Four, because the full‐text search function does not serve as a substitute for the books that are being searched.” It’s irrelevant that the libraries might be willing to purchase licenses, if their use were otherwise deemed unfair. “Lost licensing revenue counts under Factor Four only when the use serves as a substitute for the original and the full‐text‐search use does not.”
The court also rejected plaintiffs’ other, less globally significant theory of market harm: the risk of a security breach that might put their works out in the open. But the record showed extensive security measures to safeguard against that risk. This was too speculative to count as harm, though the court cautioned that it wasn’t foreclosing “a future claim based on circumstances not now predictable, and based on a different record.”
Access for print-disabled patrons: Expanded access wasn’t transformative, since the authors write books to be read (or listened to), and the HDL simply enabled a larger audience to read the works, consistent with the author’s original purpose.
Weirdness alert: The HDL’s reformatting “appears, at first glance, to be creating derivative works over which the author ordinarily maintains control.” Sure, this adaptation into a different medium is the only way the print-disabled audience can obtain access, but that’s also true for non-English speakers and books written in English.
Comment: Noooooo! Formatting isn’t a derivative work because it doesn’t add new creativity. Translation, as into ASL, would create a derivative work. But unless I completely misapprehend the techniques used for print-disabled audiences, changing text into Braille, making its size larger, changing its color, etc. no more create a derivative work than the change of a print book into an ebook creates a derivative work, even though the ebook is now made of ones and zeroes. Nor, I think, would using a substitution cipher on a book create a derivative work. Unlike the British, we don’t have a quasi-copyright right protecting typographical arrangements. Can we all agree to pretend that the court invoked the reproduction right instead? Thanks.
Anyhow, transformative use isn’t absolutely necessary for fair use. Providing access to the print-disabled is itself a valid purpose under factor one. The Supreme Court said so in Sony: “Making a copy of a copyrighted work for the convenience of a blind person is expressly identified by the House Committee Report as an example of fair use, with no suggestion that anything more than a purpose to entertain or to inform need motivate the copying.” The 1976 Act legislative history on which the Court relied expressly stated that making copies accessible “for the use of blind persons” posed a “special instance illustrating the application of the fair use doctrine . . . .” That history noted that publishers don’t usually make accessible formats commercially available, and said that an individual who made a single copy, for free, for a blind person would be a fair use. Since then, the ADA reaffirmed Congress’s commitment to “ameliorating the hardships faced by the blind and the print disabled,” as did the Chafee Amendment.
Factor two weighed against fair use, but didn’t matter. Nor was retaining entire digital image files excessive. While text files allowed text searching and text-to-speech, “the image files will provide an additional and often more useful method by which many disabled patrons, especially students and scholars, can obtain access to these works. These image files contain information, such as pictures, charts, diagrams, and the layout of the text on the printed page that cannot be converted to text or speech. None of this is captured by the HDL’s text‐only copies.” The court noted that many legally blind patrons could view these images with sufficient magnification or increased color contrast, and other patrons who couldn’t turn pages or hold books could use assistive devices to view all the content in image files. For them, access to the image files was necessary to perceive the books fully, making the retention of copies reasonable.
Factor four: it was undisputed that the present‐day market for books accessible to the print-disabled “is so insignificant that ‘it is common practice in the publishing industry for authors to forgo royalties that are generated through the sale of books manufactured in specialized formats for the blind,’” according to the appellants themselves. Comment: again, the mere desire to control something doesn’t confer a right to do so. The number of accessible books currently available is a few hundred thousand, a minute percentage of the world’s books, versus the HDL’s over ten million. Congress knew that publishers didn’t make their books available in 1976; this is still true.
Preservation copies: the HDL’s digital preservation copies ensure that books will still exist when their copyright terms lapse. As for the HDL’s propsed use to make replacement copies for a member library if (1) the member already owned an original copy, (2) the member’s original copy is lost, destroyed, or stolen, and (3) a replacement copy is unobtainable at a fair price, this wasn’t obviously ripe for resolution. The record didn’t show that the plaintiffs owned copyrights in any works that would be effectively irreplaceable at a fair price and thus would be potentially subject to being copied in case of loss or destruction of an original. Since the plaintiffs couldn’t assert the rights of others, there was no live controversy. The district court’s finding of fair use on this point was vacated and remanded for resolution of the standing issue.Likewise, the orphan works-related claims weren’t ripe. There was no indication about whether the program will be revived or what it would look like. The plaintiffs argued that the orphan works program’s legality wouldn’t depend on the specific procedures the libraries ultimately used to identify orphans, because any program resulting in the publication of complete copyrighted works would infringe. The court wasn’t persuaded. Even assuming for the sake of argument that the plaintiffs were right about infringement, it didn’t follow that any of their copyrights would be infringed. Thus, there was no “certainly impending” harm, nor any hardship if decision were withheld. If Michigan or HathiTrust reinstitutes the program in a manner that would infringe the copyrights of any proper plaintiffs, they could return to court. The mere possibility of future injury isn’t hardship.