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

Fri Aug 14 17:08:51 UTC 2026

I welcome the feedback. I am not aware of any other AI model-focused
license having been submitted previously, and I apologize if we are pushing
boundaries that OSI or this committee is not prepared to cover. I had heard
rumors there was interest within OSI in seeing the OpenMDW license
submitted, which is part of why we brought OpenMDW here rather than
treating it as outside scope. If OSI is the wrong forum, I will take that
as a fair response.

I'll try to address the questions and responses so far in one consolidated
response. Richard, I don't see the message from Luis you referred to in
your last email. I'm not sure if that is a mailing list delay or if that
email wasn't sent to the full list.

The feedback seems to focus on the termination provision. I'll repaste the
OpenMDW-1.1 [5] termination provision here so everyone can easily take
another look at it:

*If you file, maintain, or voluntarily participate in a lawsuit against
any **person or entity asserting that the Model Materials directly or
indirectly **infringe any patent or copyright, then all rights and grants
made to you **hereunder are terminated, unless that lawsuit was in response
to a **corresponding lawsuit first brought against you.*

Richard, the termination copyright trigger you focused on was a deliberate
change, and I flagged it as the sole substantive change from 1.0 for
exactly this reason. I’m including some context and responses to the
specific concerns below. I will also try to answer Pam's history question,
since the two are related. I do not know the entire history of all
licensing termination provisions, though I’m familiar with a few… I’ve
covered what I do know of here.

1. It helps to separate the termination trigger from the scope of what is
terminated.

On the remedy side, terminating copyright grants is not new for
OSI-approved licenses. A few examples:

- MPL 1.1 [1], Section 8.2(a): a patent infringement claim against a
Participant terminates, on 60 days notice and subject to a cure mechanism,
that Participant's grants under Sections 2.1 and 2.2. Each of those
sections includes the copyright grants (subsection (a)) alongside the
patent grants (subsection (b)).

- CDDL 1.0 [2], Section 6.2: a patent claim alleging the Participant
Software infringes terminates all rights granted under Sections 2.1 and
2.2, not only from the Participant but from the Initial Developer and all
Contributors.

- OSL 3.0 [3] and AFL 3.0 [4], Section 10: the entire license terminates
upon commencing a patent action against the Licensor or any licensee
alleging the Original Work infringes. The OSL’s Section 10 is below:

*10) Termination for Patent Action. This License shall terminate
automatically and You may no longer exercise any of the rights granted to
You by this License as of the date You commence an action, including a
cross-claim or counterclaim, against Licensor or any licensee alleging that
the Original Work infringes a patent. This termination provision shall not
apply for an action alleging patent infringement by combinations of the
Original Work with other software or hardware.*

So full termination of all grants, copyright included, has a history of
precedent within OSI. I suspect what is unique in OpenMDW-1.1 [5] is
copyright as a trigger for termination, and I will address that in point 3.
below.

On Pam's question about the history: my understanding is that the
retrenchment ran along two dimensions. The first was the scope of the
trigger. The IBM-PL [6] and CPL [7] provisions reached patent claims
unrelated to the licensed software (as did Facebook’s later, highly
controversial, and never-OSI-approved BSD+Patents [8] grant). I suspect
that is may be the source of the criticism Richard was also referring to. I
would agree the trend seems to have shown later licenses narrowed the
trigger to claims that the covered work itself infringes - which is what
the OpenMDW does.

The second dimension was on the remedy, with Apache-2.0 [9] in 2004
limiting termination to the patent grants alone. But that narrowing was
never universally uniform. CDDL, with full termination of all grants from
all contributors, was approved in January 2005, weeks after Apache-2.0. MPL
2.0 [10] in 2012 – and the CDDL retained termination of all Section 2.1
grants upon a patent claim, including the copyright grant in Section
2.1(a). And OSL 3.0 remains approved with whole-license termination.

The consistent point seems to be the trigger: termination should be limited
to claims about the licensed work itself. The OpenMDW-1.1 is likewise
scoped to trigger on claims against the Model Materials, or in other words,
the covered work.

2. OpenMDW-1.1’s termination trigger is based on the licensed Model
Materials (not outputs).

Three key points about the OpenMDW’s limits:

- The trigger reaches only lawsuits asserting that the Model Materials
infringe. Claims about anything other than the licensed materials do not
trigger it. This is the same formula other approved licenses align on.

- There is an express exception for suits brought in response to a suit
first brought against the licensee. The provision can only be used
defensively, not offensively. Termination is reachable only by a licensee
who initiates, unprovoked, a claim that the very materials they are using
are unlawful.

- The consequence is limited to the claimant's own license. The provision
does not bar the claim, impose a penalty, or affect any claimant who is not
exercising rights under the license. A plaintiff who is not using the Model
Materials is untouched by it.

I would also note that terminating upon suits against persons other than
the licensor has direct precedent at least in OSL 3.0 (there may be
others). I included OSL 3.0’s Section 10 above, which terminates upon a
patent action “*against Licensor or any licensee…*” Note also that OSL
3.0’s provision terminates even upon cross-claims and counterclaims.
OpenMDW-1.1, with its express exception for suits brought in defensive
response, is narrower on that axis than an already-approved license.

3. Copyright is in the trigger to maintain symmetry

The function of defensive termination has always been focused on symmetry:
a licensee should not simultaneously exercise the benefits of a license
while asserting in court that the licensed materials are unlawful. In 2004,
for software, the realistic form of that assertion was a patent claim, so
Apache-2.0 and others focused on and directly addressed patents.

Model publishers are facing a very different economic equation than circa
1990-2005 software developers. Models are built from large bodies of
existing works, and that is where the model publisher’s legal exposure
arises. For those model publishers, the realistic assertion they face is
that the licensed materials themselves infringe, and that infringement
claim likely runs under copyright, not patent law. A patent-only provision
in this context would replicate Apache-2.0's form while abandoning its
function - there would not be symmetry. And that was the clear feedback
received after OpenMDW-1.0's release.

To be clear, the OpenMDW-1.1 license does not take any position on whether
or not there may be a legitimate copyright claim against a model. The
hypothetical fact patterns will belong to the parties training and
publishing models. Anyone with a claim against the Model Materials can
bring that claim, including a licensee, and all remedies, damages, and
injunctive relief included, are unimpaired. The only consequence is that a
licensee who also contends the Model Materials are unlawful/infringing
cannot continue relying on a license to those same materials while pressing
that contention offensively. Termination puts the parties back to the
position they would be in absent the license - which brings this back to a
symmetrical position.

4. The training data hypothetical everyone’s likely thinking about.

Generative models are tools used by users to create. With enough effort and
precise, meticulous prompting, a user could hypothetically manipulate a
model to recreate what appears to be a derivative of a copyrighted work.
Even if the outputs appear infringing, that does not necessarily mean the
Model Materials were themselves infringing. Without any AI, anyone could
use Gimp today to recreate a derivative of a copyrighted work - that
doesn’t imply Gimp is itself infringing a work. For the same reason, I
think this discussion needs to separate discussion of the Model Materials
from the outputs.

A claim that a particular output infringes, without an assertion that the
Model Materials themselves infringe, does not trigger the provision. The
trigger requires a lawsuit asserting that the Model Materials infringe.
Model Materials are defined as:

*"Model Materials" means the materials provided to you under this
agreement, consisting of: (1) one or more machine learning models
(including architecture and parameters); and (2) all related artifacts
(including associated data, documentation and software) that are provided
to you hereunder.*

The termination provision’s phrase "*directly or indirectly*" parallels
Apache-2.0's “*direct or contributory patent infringement*” and addresses
theories of infringement by the Materials. It does not sweep in claims
about other works or outputs. Some open weight model licenses currently in
use in the industry (e.g. notably the Llama Community Licenses) extend
their termination triggers to claims about outputs. OpenMDW-1.1 defensive
termination does not apply to claims alleging copyright infringement of the
outputs - it does apply to claims about the Model Materials.

Where a plaintiff’s theory is that the Model Materials (e.g. the weights)
themselves are infringing, the provision applies. That plaintiff is
asserting that the licensed materials cannot lawfully be reproduced or
distributed. The license does not silence or penalize them. It says they
cannot at the same time keep exercising a license to reproduce and
distribute those same materials. That is the same symmetry rationale the
OSI accepted for patent retaliation, and I don’t see why it should be
limited to just patent retaliation.

5. The practical reality.

The reality is that we’ve seen a marked increase in bespoke model licenses
designed to cover the gaps identified in my submission. The NVIDIA Open
Model License [11] terminates on copyright or patent litigation alleging
the model infringes and was the inspiration behind the change in
OpenMDW-1.1. The current Nemotron license [12] extends its trigger to
claims that outputs infringe. The Llama family [13] (while admittedly not
OSI-approved) terminates on IP claims broadly (patent and copyright, claims
and counterclaims, outputs and weights), plus acceptable use restrictions.
These are terms important to model builders and publishers. The trend has
been that the valuable models people want to use are often not under a
license that would meet OSI’s standards.

OpenMDW was drafted to provide model builders and publishers with a
standardized license free of those restrictions: no use restrictions, no
output restrictions, and notice-only conditions. The copyright dimension of
the defensive termination provision is part of what makes a fully
permissive release viable in the current assertion environment.

6. Richard's OSD 9 question.

Treating the choice of a single license for a package of components
released together as an OSD 9 problem seems challenging to me. I do not
think OSD 9 is relevant here, honestly. OSD 9 prohibits a license from
restricting other software distributed with the licensed software.
Specifically, it says:

*The license must not place restrictions on other software that is
distributed along with the licensed software. For example, the license must
not insist that all other programs distributed on the same medium must be
open source software.*

A simple example would be a license insisting that other programs on the
same medium be open source. The Model Materials are not other software
distributed along with the licensed work. They are the licensed work –
they’re packaged and built to work together. The premise that a single
license covering a bundle of components violated OSD 9 would implicate most
software package releases I can think of. Every software distribution is a
bundle of conceptually separate components under one license: code,
documentation, build tooling, test data. The OSD has never required
per-component granularity in the grant or the remedy.

OpenMDW-1.1 places no condition on anything distributed alongside the Model
Materials. A licensee may ship them with proprietary code, copyleft code,
or anything else, and OpenMDW does not touch that software. That is the
interest OSD 9 protects. If I recall correctly, the SSPL [14] concern ran
the other way: SSPLv1 reached the service provider's own separate programs
that were never provided under the license. OpenMDW's termination reaches
only what the licensor itself provided under the agreement. IMO, the AGPL
license family is far closer to an OSD 9 issue than OpenMDW-1.1.

The symmetry rationale above also holds at the level of the Model Materials
release: a licensee suing over one component while continuing to exploit
the rest would enable both attacking and exercising the same grant.

Regarding separateness: a model release functions as a unit. The weights
are not usable without the architecture, configuration, and tokenizer, and
the documentation describes that specific release. Many apps bundle
presentation code, backend logic code, SQLite for data, and security
libraries into a release. You can swap SQLite with a pre-populated SQLite
full of numbers, and you have nearly the same package.

Thanks,

Mike

References:

[1] MPL 1.1: https://www.mozilla.org/en-US/MPL/1.1/

[2] CDDL 1.0: https://spdx.org/licenses/CDDL-1.0.html

[3] OSL 3.0: https://spdx.org/licenses/OSL-3.0.html

[4] AFL 3.0: https://spdx.org/licenses/AFL-3.0.html

[5] OpenMDW-1.1: https://openmdw.ai/license/1-1/

[6] IBM Public License 1.0: https://spdx.org/licenses/IPL-1.0.html

[7] Common Public License 1.0: https://spdx.org/licenses/CPL-1.0.html

[8] Facebook BSD+Patents (commonly referred to as the React PATENTS file):
https://github.com/facebook/react/blob/v0.12.0/PATENTS

[9] Apache-2.0: https://www.apache.org/licenses/LICENSE-2.0

[10] MPL-2.0: https://www.mozilla.org/en-US/MPL/2.0/

[11] NVIDIA Open Model License:
https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/

[12] NVIDIA Nemotron Open Model License:
https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-nemotron-open-model-license

[13] Llama 4 Community License: https://www.llama.com/llama4/license/

[14] SSPL v1:
https://www.mongodb.com/legal/licensing/server-side-public-license

[15] GPLv2: https://www.gnu.org/licenses/old-licenses/gpl-2.0.html

---
Mike Dolan
The Linux Foundation
Cell: +1.440.552.5322
mdolan at linuxfoundation.org

For help scheduling a meeting, please contact Jackline Mbithi <
jmbithi at linuxfoundation.org>.

---

On Fri, Aug 14, 2026 at 11:26 AM Richard Fontana via License-review <
license-review at lists.opensource.org> wrote:

> On Thu, Aug 13, 2026 at 1:58 PM Richard Fontana <rfontana at redhat.com>
> wrote:
>
> > Apache-2.0, MPL-2.0, and EPL-2.0 have defensive termination provisions
> > limited to patent assertions and only terminating patent licenses, if
> > I'm reading them correctly.
>
> As Luis's message shows, I was reading MPL-2.0 incorrectly. Defensive
> termination is triggered by patent assertions but terminates copyright
> as well as patent license grants. (That would seem to be true of GPLv3
> as well, unless copyright claims are implicitly also a "further
> restriction".) I think my reading of EPL-2.0 was correct.
>
> Richard
>
>
> _______________________________________________
> 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/20260814/bdc33aa7/attachment-0001.htm>