[License-review] Notice requirement for model output: OSD-compliant or not? (ModelGo)

[Moming Duan]

Hi Pamela,


Many thanks for your professional explanation. Your response is not only valuable for amending the ModelGo license but also highly educational for me. I truly appreciate your attention to detail and admire your expertise in licensing.  As an ML researcher trying to understand licensing in my own way, I recognize that GPL has an automatic licensing mechanism rather than sublicensing. Additionally, some ModelGo license variants, such as the non-commercial (NC) one, are non-sublicensable. I now see this as a flaw that neither my lawyer nor I initially identified, or perhaps they believed the notice sufficiently addressed it. I will continue consulting with legal experts to verify this matter.

Best,
Moming

> On 5 Mar 2025, at 1:43 PM, Pamela Chestek <pamela at chesteklegal.com> wrote:
> 
> On 3/4/2025 7:31 PM, Moming Duan wrote:
>> 
>> 
>>> 2. Also, by my reading, the notice requirement isn't inheritable.  So if I were to take the output from a ModelGo model, and use it to train a new model called "JoshAI", then there is no requirement that JoshAI have any particular notices.
>> 
>> Partly so. If the original model is licensed under MG-BY-OS, derivatives must also be licensed under MG-BY-OS because it is copyleft. However, for other variants like MG0, no such notice is required.  The reason I intend to add a notice in MG-BY-OS is to ensure open-source inheritance. Consider a scenario where I use an MG-BY-OS model and distribute its output as a dataset on Hugging Face without indicating which model I used. If another user downloads the dataset, trains a new model, and changes its license, they may unintentionally violate the MG-BY-OS license. This behavior is very common in current model development, and you can find many such extracted datasets on Hugging Face.
>> 
> I believe that you have a drafting flaw that means that the copyleft doesn't necessarily work, specifically that your license is sublicenseable. The GPLs are not; that is one of the brilliant aspects of the GPLs.
> 
> The BY-OS license in section 2.1(a) says "Subject to the terms and conditions of this License, the Licensor hereby grants to You a non-exclusive, non-transferable, sublicensable, irrevocable, royalty-free, worldwide right and license (including the relevant copyrights and patent rights) to ..."
> 
> The notice requirement is in 2.4(b) and says "You may Distribute the Output to third parties provided that You indicate as part of the Distribution that any Output generated through the use of the Licensed Materials and/or Derivative Materials may contain AI-generated content."
> 
> Notably in both sections the person bound is "You," defined in Section 1.1 as "you, or any other person or entity (if you are entering into this license on behalf of such person or entity and provided you have the legal authority to bind such person or entity)." I don't think that a sublicensee is necessarily "You" - perhaps you can argue that the Licensee is entering the license on behalf of its sublicensees, but I don't believe that is the intended meaning of this section. I believe the intended meaning covers the case where "You" is acting as an agent for the licensee, for example, a vendor who is creating a software program for a client. That's what I think it means in the Apache license.
> 
> So if "You" doesn't include sublicensees, then sublicensees can lawfully create Output but have no contractual obligation to label it.
> 
> There is an argument that the sublicensee might still be required to label, depending on whether Section 2.4 is considered a condition of a copyright license. But that's taking your chances that two things are true: (1) there is copyrightable subject matter and (2) the obligation to label output, in an entirely different section of the agreement without any signal language such as "on the condition that" or "provided that," would still nevertheless be a condition on the copyright license. So I wouldn't count on it.
> 
> This is a major hurdle with models, you have to assume that the obligations you want to impose related to them are only enforceable through contract, not copyright. So that means you have to find a way to be directly in privity with all users so they are contractually bound, but granting sublicenses means you aren't.
> 
> Pam
> 
> Pamela S. Chestek (in my personal capacity)
> Chestek Legal
> PLEASE NOTE OUR NEW MAILING ADDRESS
> 4641 Post St.
> Unit 4316
> El Dorado Hills, CA 95762
> +1 919-800-8033
> pamela at chesteklegal
> www.chesteklegal.com <http://www.chesteklegal.com/>
> _______________________________________________
> The opinions expressed in this email are those of the sender and not necessarily those of the Open Source Initiative. Communication from the Open Source Initiative will be sent from an opensource.org email address.
> 
> License-review mailing list
> License-review at lists.opensource.org
> http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensource.org/pipermail/license-review_lists.opensource.org/attachments/20250305/543ae76e/attachment-0001.htm>