Fair Use Week: the AI in fAIr use
The Disruptive Competition Project is taking part in Fair Use/Fair Dealing Week, an annual celebration of the important doctrines of fair use and fair dealing.
This Fair Use Week, we’re celebrating the brilliance of fair use — that it is technology-neutral, flexible, and can adapt to new technologies, as opposed to more limited specific exemptions that could quickly become obsolete. U.S. leadership in innovative fields like artificial intelligence (AI) can be attributed to policies such as fair use. As DisCo contributor Josh Landau has explained, without a properly structured environment, AI will wither on the vine.
In order to facilitate the development of AI, countries around the world are contemplating machine learning or text and data mining (TDM) exceptions to copyright law. AI relies on practices such as fair use and TDM, as large data sets are critical to train these systems. This includes generative AI and large language models (LLMs), which are trained on a wide variety of publicly available data. AI could not thrive without access to publicly available data — that’s why fair use and TDM rights are so important.
The U.S. leads the world in AI innovation, and part of this must be credited to fair use. However, the AI market is a highly competitive one and the U.S. model isn’t the only approach. There are two main approaches to enabling AI access to large data sets:
- the flexible, U.S.-style fair use approach, and
- TDM-specific legal exceptions
Unlike the U.S. approach, the more prescriptive TDM-specific legal exceptions (which exist in the European Union, Japan, Singapore, Ukraine, and the United Kingdom) may not provide the necessary legal certainty for AI development or be flexible enough to handle the necessary degrees of access to data that AI requires.
One point of note is that these models aren’t mutually exclusive; Singapore, for example, decided to adopt a specific TDM exemption in addition to fair use. Nonetheless, countries taking the latter approach generally have not seen as much AI innovation as the U.S.
Even before the AI boom, conversations about TDM exceptions have been happening globally in venues like the World Intellectual Property Organization (WIPO), where constituencies like the African Group have been advocating for TDM exemptions to support research and innovation, and in international trade agreements with countries on the Pacific Rim, which include U.S.-style fair use provisions.
The Trans-Pacific Partnership (TPP) had such a provision, drafted by the U.S. Trade Representative (USTR). The U.S. ultimately pulled out of TPP, but the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP) adopted the USTR-drafted language. Subsequently, the Regional Comprehensive Economic Partnership (RCEP) also incorporated this language. Thus, the USTR-drafted language appears in two plurilateral agreements to which the U.S. is not a party. Nonetheless, these obligations should be helpful in those countries, such as signatory Australia, where ambiguity about the scope of fair dealing might otherwise lead to restrictive policies undermining the development of AI.
Whether local companies can do the machine learning ingestion necessary for generative AI in particular may depend on the way specific statutes are worded and how courts choose to apply them in pending cases. As DisCo contributor Jonathan Band has explained, legal opinions on fair use can “provide the legal certainty necessary for the development of AI systems,” and may even “provide helpful guidance to AI developers and courts in other fair use jurisdictions, including the United States, Canada, and Korea.”
It is no surprise that flexible copyright regimes lead to further innovation. For more than a decade, CCIA research has demonstrated that the U.S.-style fair use doctrine generates immense benefits to the U.S. economy. As Band has noted, the U.S. should “not prevent other countries from adopting fair use.” Indeed, Nigeria did exactly that last year. Stakeholders in the public and private sectors have realized for many years that flexible, balanced copyright, including the fair use doctrine, is key to future-proofing copyright laws to promote new innovation and competition in the marketplace.
