Press Alt+1 for screen-reader mode, Alt+0 to cancelAccessibility Screen-Reader Guide, Feedback, and Issue Reporting | New window
All CLE Packages On Sale Now — Ends 1/16 (No Code Needed)

Generative AI and Copyright: Copyright Infringement

Presented by Yelena Ambartsumian

(767 Ratings)
LexVid

Watch This Course for Free!

New LexVid members can watch their 1st course for free. No credit card needed, just create an account and you'll receive your certificate immediately after watching your free course.

Already a member? Sign in
Video Player is loading.
Slides
move
Loaded: 0%
Current Time 0:00

Course Description

Length: 1h 0min    Published: 8/25/2025    
Part two of the Generative AI and Copyright Series continues with another one-hour CLE course that examines copyright infringement risks and legal challenges arising from generative AI systems. The session will focus on training data practices, liability for outputs, and DMCA compliance. We will analyze key litigation trends, the fair use defense to copyright infringement, and discuss actionable strategies for mitigating infringement risks in AI development and deployment.
Learning Objectives
* Understand the different architectures of Generative AI models, including Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), autoregressive models, and diffusion models
* Understand the fair use doctrine, an affirmative defense in response to copyright infringement claims
* Evaluate fair use defenses in recent AI litigation, including transformative use arguments and conflicting court interpretations
* Develop practical strategies and best practices for minimizing legal risks and advising clients on the deployment, use, and incorporation of generative AI in creative and business processes
Read the course transcript.

Speaker Q&A

Question
Do you think courts will continue to apply traditional fair use factors as-is, or will they adapt the fair use doctrine to account for AI’s unique characteristics? Although AI itself cannot infringe, can a user who "creates" a work through a series of prompts infringe (perhaps unknowingly) if the AI-generated content itself incorporates infringing material?
- PaulS
Answer
Great questions. Courts cannot change the statutory factors in the fair use test, but they can of course change how they apply them. Although the Supreme Court's recent decision in Andy Warhol Foundation v. Goldsmith clarified how to apply the first fair use factor (purpose and character of the use), by stressing the commercial aspects and explaining that the meaning or message alone does not make a use transformative, since then, we have seen a court in the Northern District of California (Kadrey) find that AI training on copyrighted materials was a fair use because the technology is new and thus highly transformative, and a second court (Bartz) find more or less the same ("quintessentially transformative" with little explanation), as far as books that were pirated (but not for books that were purchased properly, scanned, and trained on). But depending on the AI system, you are right, certain parts of the fair use test may be harder to apply. For example, for the "amount and substantiality of the potion used" factor, a diffusion model is not necessarily "remembering" what it was trained on, but rather the steps it took to noise and denoise an image. We are now seeing Midjourney's attorneys make that very argument in one of the infringement suits against Midjourney. With respect to the second question, AI can itself infringe: developers of the generative AI model or system could be liable for direct infringement or contributory or vicarious infringement, based on how the system is trained, configured, and then deployed & used by users. Similarly, someone who generated an output using the AI system could also be liable for infringement, even if they did not know about the underlying works that were used to train or create the infringing output; this is why many generative AI systems have in their terms of service or terms of use an indemnity provision for IP infringement, though there are carve outs to that that users should also be aware of. In practice, it may make more sense for the plaintiff to go after the AI system/model developer rather than the user (or go after both).
- Yelena Ambartsumian

Presented By:

Yelena Ambartsumian

New York, NY

(212) 804-5724

info@ambartlaw.com

Featured Reviews