KEI Comments and Notice of Intent to Testify Regarding 2026 Special 301 Review: Identification of Countries Under Section 182 of the Trade Act of 1974: Request for Public Comment (Docket USTR-2025-0243)
KEI_2026 Special_301_Review_Request_to_Testify_and_Comment
Introduction
Knowledge Ecology International (KEI) submits the following comments and notice of intent to testify at the hearing regarding the 2026 Special 301 Review: Identification of Countries Under Section 182 of the Trade Act of 1974: Request for Public Comment (Docket USTR-2025-0243).
As the USTR conducts its review of global IP protections and market access, KEI is concerned about unrealistic expectations regarding outcomes, given the extensive engagement of the administration on issues that offend trading partners and undermine trust and any belief that agreements with the United States will be durable and honored.
The current Administration is simultaneously applying significant pressure across a wide array of trade and political fronts, including the aggressive threat of trade restrictive tariffs, demands other countries raise drug prices, threats to annex Greenland or Canada, to rule Cuba and Venezuela, to withdraw funding of important UN agencies and programs, and a plethora of other initiatives that create an avalanche of uncertainly and resentment.
USTR has operated for years on the assumption that it is in the interest of the United States to promote high standards for intellectual property rights and to oppose exceptions and limitations to those rights. Those assumptions need to be evaluated, and are out-of-date.
Copyright and training of AI
On matters concerning copyright, governments around the world are rushing to hastily rework copyright laws to address questions about the use of copyrighted works to train a growing array of artificial intelligence services. The data used to train such services are massive, and indeed the data sets are so large they are almost impossible to comprehend.
For Large Language Models (LLMs), the term “large” is an understatement. As of 2026, the data sets used to train flagship services like Gemini, GPT-5, and Llama have reached scales that make the clearance of rights a practical impossibility.
One measure of the size of databases used to train LLMs is the Token Count (The “Vocabulary” Size). Tokens are the basic units of text (roughly 0.75 words per token). The growth in training data has been exponential:
- GPT-3 (2020): Trained on ~300 billion tokens.
- Llama 3 (2024): Trained on ~15 trillion tokens.
- Gemini 2.5 / GPT-5 (2025-2026 estimates): Modern “Frontier” models are now hitting 15 to 85 trillion tokens.
Modern services are no longer just “language” models; they are multimodal. This has exploded the dataset sizes:
- Image Data: Models like Gemini 2.0 were trained on billions of images (e.g., 2.1 billion images).
- Video Data: Current frontier models are rumored to have digested millions of hours of video (Sora and Gemini 2.0 reportedly used over 1 million hours of YouTube video).
- Audio Data: Hundreds of thousands of hours of speech and music.
AI services are typically designed to service global markets, but national laws on the legality of using copyrighted works to train AI are diverse and evolving.
The United States has a profound interest in legal regimes that are compatible with the development and improvement of AI services. This includes legal regimes in the United States, but also in other countries, where the AI services will operate.
The United States is rapidly burning bridges with its trading parties through a flood of unpredictable, trade restrictive and politically disruptive actions that make it increasingly difficult for the United States to provide leadership and guidance on global norms for copyright exceptions and limitations for AI. This is not some minor issue.
Pricing of Drugs and other Medical Technologies, and trade related aspects of biomedical R&D
President Trump has made a series of statements and executive orders that require trade officials to promote higher drug prices in foreign markets, as if this somehow is in the interest of the United States. It’s important to note that not all drugs or gene or cell therapies are owned by US companies, and that the US is going to have to find ways to lower prices, a lot, to deal with our aging population.
A better approach is to pressure foreign trading partners to increase public sector funding of biomedical research, much of it resulting in knowledge that is shared as a global public good. The US, historically, has been the largest provider of publicly funded biomedical research, including the knowledge that is shared as a global public good.
Rather than asking foreign governments to raise drug prices—an approach that predictably creates new barriers to access and is politically unpopular everywhere—the United States should insist that trading partners increase public sector funding on biomedical research, including research that has a global public goods character.
Sincerely,
James Packard Love
Knowledge Ecology International (KEI)
james.love@keionline.org
Claire Cassedy
Claire.Cassedy@keionline.org
January 28, 2026