What I Learned at the Art Law Institute
I attended the 17th annual Art Law Institute in New York back in May. The closing AI panel at the 17th Art Law Institute opened with a slide that said "not much is new." Here's why. The conversation about AI and art tends to cycle through alarm: new technology, old law, nothing figured out. What the panels made clear is that the law is settling faster than the discourse about it. The Copyright Office has issued three reports on AI. The D.C. Circuit affirmed in early 2025 that purely AI-generated works cannot be copyrighted. The framework isn't complete, but the direction is clear enough to act on.
I attended partly because I represent a working artist and was interested in picking up practical signals (and let’s face it, also because I needed CLE credits!). The day covered more than AI: a session on museum security sparked by the Louvre heist was a useful reminder of how analog many art world vulnerabilities remain, a panel on dispute resolution made a strong case for why mediation is often the right tool in art conflicts where relationships and reputation matter as much as legal rights, and a session on the Artist's Corporation model raised genuinely interesting questions about whether LLCs and sole proprietorships are actually built for how artists work today. But the AI panels were what I went for. Here's what I came away with.
Document your process if you use AI as a tool.
The law protects human creative expression, including work made with AI, if the human exercised real creative control. The Copyright Office's test isn't whether AI was involved, but whether meaningful human choices shaped the output. The clearest example from the panel materials: an artist who used Stable Diffusion but completed over 35 rounds of edits, kept a time-lapse record of the process, and submitted a detailed authorship statement ultimately got their work registered. The person who types a prompt and downloads the result gets nothing. If you use AI in your creative process, you should start building a record of how you work now.
Your moral rights die with you, and that matters more than it used to.
Under the Visual Artists Rights Act of 1990 (VARA), visual artists have rights of attribution and integrity. These are real rights with real teeth, as evidence by the $6.75 million award in the 5Pointz decision. But they terminate at the artist's death. As generative AI makes it easier to produce work in a deceased artist's style or voice, the gap that VARA's expiration creates becomes a significant vulnerability. Estate planning for artists isn't just about asset distribution. It's about what happens to your archive and your likeness when you're no longer here to control them. That planning needs to happen while you're alive and while VARA's protections are still operative.
Training on your work is a legal gray area that's getting less gray.
The Copyright Office's Part 3 report on generative AI training (which was in pre-publication at the time of the conference) pushes back on the claim that AI training is inherently fair use just because it "mimics human learning." The Office's position is that if training outputs substitute for the original work, that weighs against fair use. The Andersen v. Stability AI class action is the leading case to watch on training-data liability. For a working artist, this means the question of whether someone can scrape your work to train a model is a legal question with some developing leverage behind it, not just an ethical complaint.
I attend conferences like this because privacy, AI governance, and intellectual property intersect in ways that matter across my practice. For working artists specifically, the practical questions (namely how do I protect what I create, how do I plan for my estate, what can I do if someone trains on my work) are exactly the questions I work through with clients.
If you're a working artist and any of this sounds familiar, I'm happy to talk.
— James
