About the Authors

  • The Authors and Contributors of "Patent Docs" are patent attorneys and agents, many of whom hold doctorates in a diverse array of disciplines.
2018 Juristant Badge - MBHB_165
Juristat #4 Overall Rank

E-mail Newsletter

  • Enter your e-mail address below to receive the "Patent Docs" e-mail newsletter.

Contact the Docs


  • "Patent Docs" does not contain any legal advice whatsoever. This weblog is for informational purposes only, and its publication does not create an attorney-client relationship. In addition, nothing on "Patent Docs" constitutes a solicitation for business. This weblog is intended primarily for other attorneys. Moreover, "Patent Docs" is the personal weblog of the Authors; it is not edited by the Authors' employers or clients and, as such, no part of this weblog may be so attributed. All posts on "Patent Docs" should be double-checked for their accuracy and current applicability.
Juristat #8 Overall Rank


« International Trade Commission Issues Report of COVID-19 IP Waiver | Main | Medtronic, Inc. v. Teleflex Life Sciences Ltd. (Fed. Cir. 2023) »

November 12, 2023


I'll reserve judgment about unjust enrichment and other non-copyright claims, but at least under copyright (and putting aside any incidental copying as part of the training process), when #1 is posed as the question of whether mere training on copyrighted works infringing, the answer is very straightforward: No. It's no different than if, as an aspiring fantasy author, I started my writing preparations by studying up on every JK Rowling book. Would the mere act of my studying infringe JK Rowling's copyrights? The answer is clearly no. I don't see why that logic wouldn't apply to the LLM training context either.

Likewise, and continuing the JK Rowling hypo, the answer to #2 is also straightforwardly "Yes". If, after completing my preparations, I produce something that's substantially similar to a JK Rowling work, then of course that was an infringement.

Maybe there is something fundamentally unfair about how LLMs use authors' works on a vast scale, and again, I'll reserve comment on that, but not every fundamental unfairness that also happens to involve copyrighted works necessarily results in copyright infringement.


Eliminating the need - if Fair Use is found at the front end - has a very different set of actors (it would be the user of the AI engine, not the builder of that engine).

Divide and Conquer may be an apt analogy.

The comments to this entry are closed.

May 2024

Sun Mon Tue Wed Thu Fri Sat
      1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31