Fair Use: A Uniquely American Concept That Should Not be Exported
Recently, the Government of the United Kingdom closed the comment period for its Copyright and AI Consultation, during which it proposed an unwarranted AI copyright exception (yet again). The proposal is problematic for a host of reasons (which you can read about in the Copyright Alliance’s submission). But one particularly problematic characterization stemmed from the UK Government pointing to U.S. fair use law in its survey of international regimes with AI exceptions, suggesting unauthorized AI ingestion of copyrighted works could be broadly permissible under U.S. fair use law.
The problem is that there is no AI copyright exception under U.S. fair use law. Fair use is a fact-specific analysis from statute which is rooted in case law. In fact, there are dozens of federal cases currently pending in which the fair use question is front and center, and each case has specific facts that could lead to varied outcomes. As we close out Fair Use Week, it’s worth examining how this uniquely American doctrine has come up in the international discussion on AI and copyright issues, and how fair use law is a poor basis upon which to justify proposals and calls for AI copyright exceptions.
U.S. Fair Use Law Is Too Fact-Specific and Nuanced to Justify Broad Applications
Unfortunately, the UK Government is not the first or only country to point to U.S. fair use law as part of its justification for introducing a harmful and broad AI copyright exception that would decimate the creative economy. In its AI and Copyright Consultation, the Hong Kong Government claimed that its proposal of a broad statutory AI copyright exception was justified in their own region because AI exceptions may already be permitted in other countries, including in the U.S. under fair use. As we pointed out in our comments to the Hong Kong Government, this is an overbroad characterization of U.S. fair use law, and it disregards the nuanced nature of the doctrine as established in over a century’s worth of cases.
Though fair use and its recognizable four factor test is enumerated in the provisions of section 107 of the Copyright Act, its nuanced application continues to be shaped by decisions from U.S. courts, where judges actively interpret and apply the statutory factors in different, fact-specific contexts. As a result, the exception’s application and impact continues to evolve. That’s nowhere more apparent than with the U.S. Supreme Court recent decision in AWF v. Goldsmith case, which recalibrated the boundaries of the factor-one analysis.
This is a key quality of fair use that courts take time to repeatedly stress in fair use-related opinions: that the exception is a uniquely fact-intensive and case-specific inquiry and thus, there can be no categorical exemptions under the doctrine. That is not to say that helpful precedent can’t be set based on closely analogous cases. But the reality is that because no two cases will be exactly the same in their fact patterns, it is a mistaken characterization (especially in the context of discussing a more permanent statutory copyright exception) that infringing uses could be categorically exempted under U.S. fair use principles. In a prior blog post, we previously explored some cases that AI companies wrongly cite to for the proposition that AI training is fair use and how those cases highlight the context-specific nature of fair use. If these fair use cases prove anything, it is that AI presents novel questions that are not analogous to cases involving prior technology like VCRs. Courts are already recognizing that the AI cases today are unlike that of prior copyright and technology cases that have been litigated.
Yet, as mentioned before, the U.S. fair use exception seems to be a perennial darling for foreign governments that are looking to justify introducing AI copyright exceptions. But in reality, the fair use doctrine’s complex application in fact-specific contexts makes it an ineffective foundation for any country to build sweeping copyright exceptions upon— exceptions which would indiscriminately and broadly affect every creator and copyright owner in the creative industries.
A U.S. Court Has Already Rejected Fair Use Arguments
As the nearly forty AI lawsuits concerning unauthorized ingestion of copyrighted works continue to move through the federal courts, it is becoming more apparent that AI companies are losing ground in their legal arguments and that courts are closely scrutinizing the evidence over AI ingestion issues. As we highlight in this blog post describing AI litigation developments during 2024 in the U.S. and this blog post describing some key takeaways in a few key AI cases, the preliminary wins in these cases demonstrate that there are no clear-cut answers to the fair use question—and certainly no clear-cut AI fair use exemption.
This is no more apparent than in the first fair use decision in an AI-related case, Thomson Reuters v. Ross Intelligence, decided earlier this month. In that case, a federal district court in Delaware found that defendant’s copying of legal headnotes to train an AI model did not qualify for the fair use exception (we further explore the impacts and analysis of that decision in this blog post). Despite the court noting that the case did not involve generative AI, the legal analysis of the Ross decision, coupled with the recent Supreme Court decision in AWF v. Goldsmith demonstrates that the AI companies’ position that unauthorized training on copyrighted works is permitted by fair use is not nearly as solid as these companies first believed. In fact, that is probably why we are seeing more license agreements being reached between AI companies and rights holders.
To be clear, courts will continue to weigh in on fair use arguments in other AI and copyright cases this year. But these early developments serve as an important reminder and a warning to foreign countries, of the fact-specific nature of fair use law and the error of relying on or even pointing to US fair use law as a basis for creating an AI exception in their own copyright law.
Conclusion
As the global AI race continues, there will continue to be efforts to find shortcuts in the name of progress—shortcuts that inevitably harm the creative community. These efforts include shortcuts by way of introducing AI copyright exceptions, including those that are wrongly justified on shaky foundations like U.S. fair use law. These approaches are ill-advised, not only because they lack sound foundational, legal basis, but also because such shortcuts inevitably lead to a global race to the bottom.
If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!