Two weeks, two copyright conferences, & a whole lotta AI and ethics!

Late May and early June were quite busy for me, as I hosted and then attended two conferences two weeks apart. Both conferences were focused on copyright, and both had a lot about AI and ethics (sometimes related, sometimes not). Rather than report on each conference separately, I’ve grouped my observations below around those two main themes.

I was excited to host the 2024 UIPO Annual Meeting at Wake Forest this year, May 29-31. UIPO (University Information Policy Officers) members’ libraries and institutions serve as hosts, underwriting the costs of the conference so that we don’t have to charge a registration fee. Many thanks to Dean Tim Pyatt for his support in making this happen. We had 26 in person and 36 online attendees, with a mix of presenters on site and online. Huge kudos to Barry Davis for ensuring all of the tech ran smoothly. Over the course of two days, we delved into various copyright issues, from case updates to ethics to structural inequalities to AI (of course). We wrapped up the conference with a fun fair use trivia game that managed to stump even the most seasoned lawyer-librarians in the room!

Eleven days later I found myself in Colorado Springs for the annual Kraemer Copyright Conference at the University of Colorado Colorado Springs, June 11-13. I’ve attended this conference in the past and it was great to return after a pandemic hiatus. I saw several colleagues from UIPO at Kraemer, including some who’d been unable to come to our meeting at Wake. While the primary focus of this year’s conference was AI, there were a few sessions that touched on other aspects of copyright. A benefit of this conference is that it pulls in copyright experts who are not librarians – law professors, practicing attorneys, and policy wonks – so attendees get a broader perspective on copyright issues.

I’ll be happy to sit down over coffee to share more details if you’re interested!

Copyright & AI

• There are legal issues on both sides – input & output – and copyright may not be the most important issue we should be talking about, although it is important and raises three big questions:
  • Is copying for training data infringing?
  • Are AI-generated outputs infringing?
  • Are AI-generated outputs copyrightable?
    • This raises questions about what it means to be an author
• Looks like individual users will be blamed for inappropriate AI uses, which lets AI companies off the hook
  • If you don’t hold the people making the LLMs (large language models) responsible, then is regulation even possible?
• Robust training data can help reduce biases (which we’re already seeing in LLMs), but questions remain about whether this is actually possible
  • Good argument for including copyrighted, recent content

Copyright, Racism, & Ethical Stewardship

• Copyright and intellectual property laws are framed as systems of ownership and power, which is purview of whiteness
• Concepts of ownership vary across cultures: indigenous cultures often have less fixed or documented traditional knowledge (e.g., oral traditions), creating disadvantages when framed in terms of “ownership” or when ownership is rejected outright
• Using Creative Commons licenses, which requiring attribution, still implies ownership so it’s not an easy solution
• Access to traditional cultural expressions may be enhanced by digital access, but communities often gain little from that access and risk losing ownership and use of their traditional knowledge because copyright and other IP laws don’t apply therefore they don’t offer protection (catch-22)
• Racial and sexist biases of those who created and influence copyright law in turn reinforce and enshrine biases
  • Works Made for Hire create a power differential and imbalance between the copyright holder and the person actually creating the content
  • Works excluded from copyright or other IP protections often reflect works that have traditionally been associated with women or underrepresented groups, such as textile designs, clothing designs, and recipes
• Conflating copyright law and ethics can disempower users
  • Defaults to a simplistic view of copyright, losing important nuance
  • Copyright & IP laws vary internationally, so if following U.S. copyright law is “good” and “ethical,” it’s therefore unethical to follow differing international IP laws – not true!
  • Using fair use while legal may not be ethical
• Our collections aren’t faceless!
  • What does it mean to put our collections out not as people but as data?
  • We need to be thinking about how these may be feeding into AI (assume they already are)
  • AI raises risks of potential harm in sharing some of our sensitive collections