This was the 15th of the highly popular Long Arm of the Law sessions at the Charleston Conference which was founded and moderated by Ann Okerson, Director of the Offline Internet Consortium. This year's presenters were Roy Kaufman, Managing Director, Copyright Clearance Center (CCC), and Nancy Kopans, VP and General Counsel, Ithaka. The theme this year was "AI in the middle of the producers and the users".

Roy began by discussing AI and copyright and noted that AI is a series of technologies and algorithms, not a use. We need to keep ethics and humans in the middle.

The purpose of copyright is to give the user exclusive rights to reproduce the work or make derivative rights. AI is transformative. and appears in Section 101 of the copyright law, so it is a derivative work. 4 factors defining fair use are

• The purpose and character of the use, including whether it is for commercial or educational use,
• The nature of the copyrighted work,
• The amount of the portion use in comparison with the whole, and
• The effect of the use on the potential market for or value of the copyrighted work.

Some examples of recent copyright decisions: In a case regarding Kinko's, they made copies of articles to sell to students. Since they are not the student, their use was deemed to be commercial. In Thomson Reuters v. Ross, surreptitious copying of the entire Westlaw database survived a motion to dismiss. There will be a trial next year. A Google Books Case determined that all training use is fair use. The Fox News v. TVEyes court said it is transformative and TVEyes is unlawfully profiting from the work of others. Warhol v. Goldsmith found that the same copying may be fair when used for one purpose but not another. Other countries have their own laws, and most of them have not addressed the use of AI in copyright.

Nancy presented a view from the uncomfortable middle. JSTOR works between end users, content providers, and libraries, colleges, and universities.

AI is not the first technology to affect the printed word. Printing started with Gutenberg, then moved to the copying machine and the rise of the internet which revolutionizes how content is consumed and used. Generative means creating new content based on the underlying training content. Some capabilities of AI:

• AI and our community: opportunities and risks.
• AI offers tremendous possibilities for libraries. It can accelerate content processing and enhance libraries' role as a source for research.
• AI transforms how end users engage with works and provides a new era of learning, improves research work processes, and enhances opportunities
• AI can help streamline the content creation process and provide data analytics for strategic decision making.

Concerns and challenges: copyright infringement, cannibalizing the market, maintaining quality of the scholarly record, and repurposing content licenses in ways the creators did not anticipate.

Key areas of conflict and opportunity. Is training an LLM copyright infringement? What are fair use defenses? What is the market harm? What are contractual restrictions and terms of service considerations for aggregators? It's not just your decision; you are a steward of the content. Did the aggregator agree to the use? There are legitimate concerns. What are ways to navigate in this case and manage the risks? We could rely on fair use which is geographically limited because it's a US law. What ways can outputs be restricted so that users can be confident that they will not be infringing? A licensing restriction: some content providers are very reluctant for their data to be used by AI. Some libraries want training LLMs for their own scholarly use on their own campuses and unlimited fair use. 

Food for thought for libraries and AI: accelerating content processing, balancing libraries' role as a source for research, concerns and challenges. Is this a time to revisit orphan works legislation?

Perhaps the view from the uncomfortable middle should become the view from the hopeful middle.

Donald T. Hawkins, Conference Blogger