If you work with contracts regarding AI, you need to become familiar with the concept of “model distillation,” and the risks it creates for AI vendors. In short:
Through model distillation (technically, “black box” distillation), one company can create a gen-AI model by “copying the work” of another – quickly and at low cost. That’s what DeepSeek (allegedly) did to ChatGPT. Most gen-AI providers need terms protecting them from model distillation. But, while terms forbidding access to the system for competitive use should block distillation, they may be unenforceable — because of copyright preemption.
Vendors need to grapple with this issue to protect themselves. And AI customers need to understand what’s troubling their vendors, in some cases leading to odd terms.
In this video, our founder, David Tollen, explains.
Interested in learning more about information technology contracts, including those involving AI? We offer a wide range of training suited for contract professionals.
The 2025 live Master Class series is over. To find out about future live Master Classes and other webinars, please see our live webinar page here. And please see here for recorded courses available now, on-demand.
And for more about copyright preemption, please see our blog post here: New Data Protection Terms.
THIS ARTICLE, AND VIDEO, ARE NOT LEGAL ADVICE. THEY ARE GENERAL IN NATURE AND MAY NOT BE SUFFICIENT FOR A SPECIFIC CONTRACTUAL, TECHNOLOGICAL, OR LEGAL PROBLEM OR DISPUTE, AND ARE NOT PROVIDED WITH ANY GUARANTEE, WARRANTY, OR REPRESENTATION. LEGAL SITUATIONS VARY, SO BEFORE ACTING ON ANY SUGGESTION IN THIS ARTICLE OR VIDEO, YOU SHOULD CONSULT A QUALIFIED ATTORNEY REGARDING YOUR SPECIFIC MATTER OR NEED.
[Updated October 22, 2025]
© 2025 Tech Contracts Academy, LLC