Annually, the Office of the United States Trade Representative spearheads the Special 301 review process, whose goal is to produce a report that names-and-shames countries that have committed perceived offenses against intellectual property rights. The USTR solicits written comments from stakeholders, holds a public hearing wherein stakeholders provide oral testimony and can then follow up on any questions from the government in writing as well. KEI has tracked this process for over 15 years and maintains a page cataloging past Special 301 reports and highlighting key comments submitted to the annual process.
KEI submitted written comments to the USTR outlining our concerns regarding IP and AI issues (available here as PDF, and here on the official docket, USTR-2024-0023), and provided testimony at the hearing on the 2025 Special 301 review, based upon the notes below.
Notes from the KEI Oral Statement at the 2025 Special 301 Hearing (docket number USTR-2024-0023)
KEI appreciates the opportunity to offer comments to the 2025 Special 301 process. We also appreciate that USTR is under new leadership and direction this year and that new policies are still taking shape, perhaps as we speak. That said, the issue of artificial intelligence and how it relates to intellectual property seems an important one to raise now.
The Presidential Executive Order of January 23, 2025, titled “Removing Barriers To American Leadership In Artificial Intelligence,” states:
Sec. 2. Policy. It is the policy of the United States to sustain and enhance America’s global AI dominance in order to promote human flourishing, economic competitiveness, and national security.
The details to implement that policy are left to be determined later, after reviews led by the Assistant to the President for Science and Technology (APST), the Special Advisor for AI and Crypto, and the Assistant to the President for National Security Affairs (APNSA), among others.
As regards AI and IP, one pressing set of issues concerns the extent to which the developers of AI services can use copyrighted materials and non-copyrighted data to train AI services. This has burst on the scene as a policy question of high importance, and state practice is not harmonized, at all. Among the policies proposed around the world are policies that would consider unauthorized use of copyrighted material to be an infringement, or subject to text and data mining exceptions. There are various statutory licensing regimes that may be of an opt-in or opt-out nature, and there is a diversity of approaches regarding whether the training or use is considered commercial or noncommercial for purposes of an exception or statutory licensing regime, among other nuances.
The United States currently is the center of much of the most significant progress in developing AI services, but the future is anything but clear, and not just because of the recent success of DeepSeek in China.
There are many intellectual property issues raised by AI, on both the input and the output sides, but in the near term, perhaps the landscape of copyright policies regarding the use of works to train AI are the most pressing.
Any country that provides robust exceptions for using copyrighted material for AI will have a significant advantage in terms of training such services. But a lack of harmonization may create a situation where services developed in one country, such as in the United States, will not be legal in another, because of non-authorized use of copyrighted works to train the service. This makes it a significant trade issue.
The most emotive and politically important voices to make it an infringement of copyright to train AI services are cultural industries and journalism. KEI has been concerned that restrictive policies on the uses of copyrighted works to train AI will be extended to much broader classes of works. A general exception should be a broad text and data mining exception for AI, but we’d be open minded about whether there should be carve-outs for certain cultural industries and journalism. What we don’t like is a one-size-fits all approach that would treat areas like cultural industries and biomedical research the same.
We believe that AI will be increasingly used in law, scientific and biomedical R&D, and in those areas, society is best off if AI services have access to everything possible in terms of data, and where omissions may cost someone their health or freedom.
In regard to scientific and medical information, we note that unlike novelists or screenwriters, research journal authors are rarely paid for their works, and a handful of companies control a large number of journals, many of them foreign-owned. It would be a very bad outcome if private equity firms were able to significantly limit which companies can use the leading medical journals to train AI services, or for that matter, if any publisher can opt-out of the science being used to train programs that are used for drug discovery or to treat and protect patients. We’re not saying that screenwriters could or could not opt-out, but that an opt-out right should be an exception to a more general exception, and not the rule.
Down the road governments may develop more forward-looking policy frameworks to address the myriad of issues concerning copyright and AI, and also access to non-copyrighted data, in ways that are equitable, respect privacy, do not limit competition to undermine innovation, and otherwise benefit and protect society.
USTR should develop policies that are not enacted in haste, ensuring that any policies do not undermine the global commercial markets for services provided by US companies, or have bad outcomes for society in the longer run. KEI would like to work on this issue with USTR going forward as it seeks to tackle this issue internationally.
KEI would also like to highlight several points from KEI’s 2024 Special 301 comments that are still relevant:
- USTR policy on the use of exceptions to exclusive rights in patents, data, biologic resources and other knowledge goods should be consistent with Paragraph 4 of the WTO Doha Declaration on TRIPS and Public Health.
- USTR should address the threats to two important copyright exceptions, the quotation right and the news of the day exception.
- Trade related aspects of funding biomedical R&D should focus less on intellectual property norms and more on the direct and indirect funding of research by the public sector.
- Trade related aspects of public goods continue to be a neglected area of trade policy.