Academic publishers and universities are confronting a pivotal decision about artificial intelligence and intellectual property rights, one that will shape how research is used to train AI systems for years to come. As lawsuits over AI training data pile up, institutions must decide whether to accept settlements, enter licensing agreements with AI companies, or fight to protect their copyrighted works. The stakes extend far beyond money; they involve questions about the future of human creativity, scholarly integrity, and what it means for AI to truly transform knowledge.

What Makes AI Training on Academic Work Different From Fair Use?

The concept of "transformation" sits at the heart of this debate. Under U.S. copyright law, uses of copyrighted material are more likely to be considered fair use if they are "transformative," meaning they add something new with a different purpose or character and do not substitute for the original work. However, the author of a major analysis on this issue argues that there is a crucial distinction between legal transformation and the kind of meaningful transformation that depends on uniquely human capacities for thinking, feeling, and complex analysis.

Large language models, or LLMs, are AI systems trained on vast amounts of text data to predict and generate human-like language. When companies train these models on copyrighted academic articles, books, and other published works without permission, they argue the use is transformative because the model learns patterns rather than copying text directly. Yet evidence suggests that LLMs sometimes retain and reproduce unaltered portions of the works they were trained on, raising questions about whether this truly qualifies as fair use under the law.

The American Association of University Professors, or AAUP, has been examining these tensions closely. In a July 2025 report titled "Artificial Intelligence and Academic Professions," an ad hoc AAUP committee issued recommendations addressing members' concerns about the impact of AI on campus and the role of academic workers in institutional decision-making related to educational technology, including intellectual property rights over course materials.

How Should Academic Institutions Approach AI Copyright Decisions?

• Evaluate Settlement Offers Carefully: When AI companies offer to settle lawsuits or pay licensing fees, institutions should recognize that these modest financial gains facilitate the disproportionate enrichment of AI companies that use copyrighted material for training. The money flowing to publishers and universities is small compared to the value AI companies extract from that content.
• Protect Intellectual Property Standards: Academic journals and university presses must uphold ethical standards and principles of copyright law by committing to publish human-authored works and preventing the misuse of AI in academic research and writing. This means establishing clear editorial policies about how AI can and cannot be used.
• Maintain Scholarly Integrity Over Profit: Some individuals and organizations will encounter opportunities to "cash in" on AI licensing deals. However, prioritizing modest financial gains over the long-term integrity of academic publishing could undermine the credibility and autonomy of scholarly institutions.

The AAUP's 1999 Statement on Copyright affirms the rights of faculty members with respect to their institutions, establishing a foundation for how academic workers should maintain control over their intellectual property. As AI becomes impossible to ignore in academic settings, those who value scholarly and creative endeavors are increasingly thinking about what human brains can do that machines cannot, despite the tech industry's aspirations.

One key concern involves the pattern-based predictive logic of AI systems. While companies like Palantir display slogans such as "AI-powered automation for every decision" on their websites, this approach cannot replicate the kind of sensible and sensitive personalized guidance that comes from human expertise. In fields like medicine, for example, AI overviews in search engines can add confusion rather than clarity. What patients and researchers need is the transformative knowledge and insight of experts familiar with individual cases and the broader context of published research.

The emotionally nuanced creative transformation that characterizes great academic and artistic works seems far more elusive for AI systems. LLMs lack the deeply rooted cultural context of socially acquired language, common sense, and human capacities for caring, sharing concerns and commitments, and experiencing embodied emotions and individual and collective moods. This gap matters not just for creative writing but for intellectual integrity and factual accuracy in academic publishing.

As authors, editors, and publishers navigate these decisions, they will need to make consequential intellectual property choices. The outcomes of these choices will ripple across academia and beyond, shaping whether future generations of scholars can maintain autonomy over their work or whether that autonomy becomes another resource extracted by the AI industry.