AI Law & Policy Update: Copyright Lawsuits Escalate and New AI bills Advance
Artists overcome a motion to dismiss in their lawsuit against image generation systems, new AI Laws are proposed at the federal and state level, and Colorado's AI Act is ready to be signed into law.
New AI Bills and Considered Rule Changes
US Senators Mark Warner & Thom Tillis introduce the Secure AI Act of 2024, one of many newly proposed AI bills at the federal level.
The debate heats up over California’s SB-1047 Bill (The Safe and Secure Innovation for Frontier Artificial Intelligence Models Act), as the bill comes closer to passing.
The Colorado AI Act has passed and will likely be signed into law by the Governor. It includes having a right to appeal AI decisions in high-risk settings and other consumer protection requirements for those that could potential be impacted by AI.
I testified before the Advisory Committee on Evidence Rules for the US Federal Courts. I’ll be posting a more formal write-up of what I said soon, but I generally agreed that the Federal Rules of Evidence don’t need an update yet when it comes to identifying “fake” evidence generated by AI. However, an update may be needed to more thoroughly vet studies assisted by AI (e.g., language models to interpret the meaning of language or facial recognition to prove someone is in a video).
AI Copyright Roundup
A flurry of new AI copyright lawsuits are filed against NVIDIA, Mosiac, Databricks, Google, and OpenAI.
The lawsuit against Stability AI, Midjourney, and DeviantArt moves to discovery.
Only the DMCA claims were dismissed, where the court stated that it would follow a recent Doe 1 v. GitHub decision and require identical outputs for a DMCA 1202 claim.
The direct/indirect infringement claims survive, with the court saying, “Plaintiffs have plausibly alleged facts to suggest compress copies, or effective compressed copies albeit stored as mathematical information, of their works are contained in the versions of Stable Diffusion identified.”
A new Supreme Court decision, in my opinion, quashes any slim hope that OpenAI and other AI companies could mount a successful 507(b) argument that infringement occurred 3+ years ago and thus is past the statute of limitations. So much for training a model and sitting on it for three years to get around the direct infringement claim!
The Copyright Office finally accepts registration for an AI-assisted book. The caveat that it is only a compilation right. This might be a middle ground solution for many AI-generated pieces where humans do significant work in compiling various AI-generated components.
New Guidance on AI
The Dutch Data Protection Authority issued a report suggesting that scraping is almost always illegal under the GDPR.
OECD revises its AI principles in light of general purpose systems like foundation models, including new guidance on handling misinformation/disinformation concerns.
A number of researchers, myself included, continue raising concerns about the increasing likelihood that AI will be widely integrated into warfare due to global competition concerns. See pieces here, here, and here.
Who am I? I’m an Assistant Professor at Princeton University with appointments in the Department of Computer Science and the School of Public & International Affairs. Previously I received a JD-PhD from Stanford University. You can learn more about my research here. Every once in a while, I round up news at the intersection of Law, Policy, and AI. Feel free to send me things that you think should be highlighted @PeterHndrsn. Also… just in case, none of this is legal advice, and any views I express here are purely my own and are not those of any entity, organization, government, or other person.