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

[Richard Fontana]

I know it's a trope (possibly first articulated by me?) that *any*
purported open source license can be characterized as an OSD 5/6
violation, but I think it's useful to consider how this license might
be improperly discriminatory (in addition to the possible OSD 9 issue
I've raised separately) if only as a thought exercise.

This license can be seen as discriminating against a class of
potential licensees that include *open source developers*, that is to
say, human licensors of OSI-approved licenses applied to code, who
might seek to bring claims enforcing their rights against the releaser
of an OpenMDW-1.1-licensed LLM on the theory that the model or its
output breaches the OSI-approved license. I feel that gets at what is
troubling to me about this defensive termination provision. If the OSI
were to approve this license, wouldn't the perceived message be that
OSI is siding against licensors of OSI-approved licenses seeking to
pursue what those licensors believe are legitimate claims of copyright
infringement (and/or open source license breach)? As far as I am aware
no previous license submitted for OSI approval has raised an issue
remotely similar to this.

The counterpart issue never came up with patent assertion defensive
termination, for the obvious reason that offensive use of software
patents was widely regarded as bad, or at least highly problematic, in
the open source community. Indeed, software patents as a basic policy
problem was something that unified individual and corporate
participants in the early post-1998 open source era, to the point
where by 2005 the FSF was probably looking a bit awkward in not having
a license that contained what they called a "patent retaliation"
clause (so far as I can recall, the FSF kind of stood alone at that
time in trying to make the case that there was something problematic
about such provisions, at least from a GPL-oriented standpoint). While
tactics and strategies around enforcement of open source licenses has
certainly been a controversial issue in open source, as a baseline
matter there has never been a view in the general community that
good-faith enforcement of open source licenses through pursuit of
copyright claims or similar is bad or problematic per se.

To paraphrase a line from one of my favorite movies, I'm not telling
you what *I* think, I'm telling you what large numbers of open source
community members are likely to think.

Richard

On Sat, Aug 15, 2026 at 8:53 PM Kevin P. Fleming via License-review
<license-review at lists.opensource.org> wrote:
>
> I posted this on social media, but might as well post it here too even though IANAL and this could be wildly pointless, but...
>
> Licenses which provide for copyright termination as the result of a recipient suing (or otherwise engaging) the licensor have one very significant difference from this one: the content being distributed under the license can be directly inspected by anyone who has a copy to determine the validity of a copyright infringement claim.
>
> Works, particularly models, distributed under OpenMDW do not have that attribute. If a recipient believes that the model infringes on their copyrights, the only way that can be 'proven' (for some definition of proven) is to execute the model on a computer and feed it inputs in order to generate the claimed infringing output. The infringement, or lack thereof, cannot be determined purely by inspecting the model.
>
> The result of this could be that a party who claims the model infringes on their copyright and takes action against the model's licensor cannot participate in generating evidence of the claimed infringement after they have taken the action, because they no longer have permission to make use of the model in any capacity. If the process of trying their claim requires additional supporting evidence to be provided, that will have to be generated by a third party who is not an actor in the claim itself (presumably any attorneys or other representatives hired by the claimant would also be unable to execute the model since they are participants in the action against the licensor).