[License-review] For Approval: OpenMDW License Agreement, versions 1.1 (OpenMDW-1.1)

[Richard Fontana]

On Sat, Aug 29, 2026 at 8:46 PM McCoy Smith <mccoy at lexpan.law> wrote:
>
> This seems to me to be primarily a criticism of the manner in which this
> particular licensor has expressed the license or licenses for the
> materials that they have posted. I'm not sure I understand how this
> particular situation is an implication of the license text itself.

NVIDIA seems to be in some sense the 'poster child' for adoption of
this license. So I think any use of this license by NVIDIA is
especially significant and instructive.

Anyway my concern is that the example I gave illustrates a general
problem that is going to arise with use of this license, because of
prevailing practices in the AI model publisher community. We can
expect that repositories of stuff stated to be under OpenMDW will
contain all sorts of things that will be difficult to fit into the
notion that there is one clear set of "Model Materials", which in turn
will make it difficult to understand the reach of the termination
provision. This is the core feature of the license and ironically may
be its basic practical flaw.

To your next point, I recognize that I might be identifying a problem
that has been latent in widely used open source licenses all along. I
think the key concept of "Model Materials" however is what's
distinctive - it's designed to sweep in a categorically broad set of
items, in a way that I don't think is true of conventional open source
licenses.

> I will note that the ModelGo licenses (also directed to AI models) that
> were submitted and are currently up for consideration articulate the
> licensed materials in this way too:
>
> https://github.com/Xtra-Computing/ModelGo/blob/main/MGL/V2/MG0/LICENSE
> (""Licensed Materials" means the Model and Complementary Materials that
> are Distributed by the Licensor under this License. ")
>
> https://github.com/Xtra-Computing/ModelGo/blob/main/MGL/V2/MG-BY/LICENSE
> (""Licensed Materials" means the Model and Complementary Materials that
> are Distributed by the Licensor under this License.")
>
> So I'd like to hear if a similar concern about that definition is a
> problem for those licenses as well. And if not, why not.

To be honest I haven't paid much attention to the ModelGo licenses
(sorry Moming Duan) because they aren't seeing any notable use in the
community as far as I'm aware, and because the prospect of OSI
approval seems at present to be unlikely without further significant
revisions, based on the feedback I have half-followed on
license-review. OpenMDW-1.0 and OpenMDW-1.1 are already seeing
significant use, as Mike demonstrated, including by some of the most
prominent companies and labs in the AI space.

But looking at the ModelGo licenses now, I see that at least the
"sharealike" variant seems to raise the same sort of issue because of
the "Complimentary Materials" concept.

I do wonder what would have happened if Moming Duan had put a
copyright-lawsuit-triggered termination in the sharealike ModelGo
license before the publication of OpenMDW-1.1.

Richard