FAQ

The OpenMDW License Agreement v1.1 (“OpenMDW‑1.1”) is a permissive open‑source license crafted specifically for machine‑learning models and their associated artifacts (“Model Materials”). Its goal is to provide a single, clear framework that providers of Model Materials can apply to grant rights to downstream users under all relevant intellectual‑property regimes, including copyright, patent, database and trade secrets.

OpenMDW-1.1 is a minor release, with one change as compared to OpenMDW-1.0: the change from “patent” to “patent or copyright” in the fifth paragraph. See question 6 below for more details.

Under OpenMDW‑1.1, Model Materials include:

1. Machine‑learning models (architecture and parameters); and
2. All related artifacts (including associated data, documentation and software) that are provided under OpenMDW-1.1.

Item 1 is intended to include model components such as architecture definitions and learned parameters. Item 2 is intended to include related artifacts such as training or evaluation datasets, code (preprocessing, inference, training scripts), documentation, metadata and supporting tools; in each case for item 2, when the model provider also makes them available under OpenMDW-1.1.

This broad definition enables components that are critical to reproducing or extending a model to be licensed consistently. While OpenMDW-1.1 does not mandate completeness (in the sense of requiring all related components to be provided), it does provide terms that ensure the ability to use in an open and free manner any and all components that are distributed under OpenMDW-1.1.

Subject to compliance with its terms, OpenMDW‑1.1 grants unrestricted, royalty‑free permission to “deal in the Model Materials without restriction, including under all copyright, patent, database, and trade secret rights included or embodied therein.”

This is intended to include activities such as:

• Use (run or test) the Model Materials
• Copy and modify them
• Distribute originals or modifications

The license imposes no field‑of‑use, royalty or geographic restrictions, encouraging broad adoption and innovation.

Applicable laws might impose their own requirements or restrictions, which of course no license is able to override. But OpenMDW-1.1 does not impose any such restrictions of its own.

When distributing any portion of the Model Materials—whether verbatim or modified—you must retain in your distribution:

1. A copy of the OpenMDW‑1.1 license text (the LICENSE file).
2. All original copyright and origin notices present in the Model Materials that apply to the portion you’re distributing.

No further “copyleft” or “share‑alike” requirements are imposed by OpenMDW-1.1.

The OpenMDW license can be used with the MOF simply by putting the LICENSE file into the root of the repository for the model distribution. The OpenMDW does not require that components be included, but for those components that are included (Model Materials) the OpenMDW license applies to them if not otherwise provided under a different license.  Classification with the MOF is entirely separate and is led by the Generative AI Commons community who maintain the Model Openness Leaderboard and Tool at https://isitopen.ai

OpenMDW‑1.1 includes a patent and copyright litigation‑termination provision: if you initiate or maintain a lawsuit alleging that the Model Materials (directly or indirectly) infringe a patent or copyright—except in defensive response to a suit first filed against you—all rights granted to you under OpenMDW‑1.1 terminate immediately.

Although the scope is broader, this provision is intended to be comparable to “defensive termination” provisions relating to patents in other licenses such as Apache-2.0. Defensive termination provisions cannot ensure the absence of patent aggression, such as from “non-practicing entities” or other patent holders that don’t make use of the OpenMDW-1.1 materials themselves. However, this provision aims to discourage aggressive patent or copyright assertions from among the users of the Model Materials that could hinder open collaboration.

No. For any outputs you generate by using the Model Materials, no restrictions or obligations are imposed by OpenMDW‑1.1 for using, modifying or sharing those outputs.

This is intended to give comfort that outputs you produce by running or fine‑tuning the Model Materials—such as generated text, images or predictions—are not subject to restrictions imposed by the model providers. Although applicable laws may impose restrictions or obligations, OpenMDW itself does not restrict your rights to use, modify or commercialize those outputs as you see fit.

The Model Materials are provided “AS IS”, without any warranty—express or implied—including but not limited to warranties of merchantability, fitness for a particular purpose, accuracy or non‑infringement. The license also disclaims all liability for damages arising from use of the materials, to the fullest extent permitted by law

The disclaimers at the end of the license include several specific provisions disclaiming responsibilities as to the contents of the Model Materials and your use of them. You are solely responsible for, and the providers of the Model Materials disclaim responsibility and liability for, clearing any applicable third-party rights applicable to your use of the Model Materials.

1. Copy the full text of the OpenMDW‑1.1 license into a file named LICENSE at the root of your repository.
2. Ensure all Model Materials (models, code, data, etc.) that you want to distribute under OpenMDW-1.1 are contained in that repository.
3. Include copyright notices in each file or artifact that you want downstream users to preserve.
4. Include any additional documentation (e.g. README.md) describing your project, and referencing the OpenMDW-1.1 license.
5. If your repository includes third-party materials that you received under a non-OpenMDW license, include copies of those license texts and clearly explain which portions those licenses apply to.

In many cases, yes, we believe so. “License compatibility” is a complex and fact-specific topic. But in many cases (especially those involving other “permissive” open source licenses) OpenMDW-1.1 is likely to be compatible, because it is possible to comply with both of them at the same time.

Because it is permissive and imposes minimal conditions, you can generally combine Model Materials licensed under OpenMDW‑1.1 with code under MIT, Apache-2.0, the BSD family of licenses, and similarly permissive licenses. Care should be taken if integrating with copyleft licenses (e.g. GPL-3.0), as their terms may impose additional requirements on the combined work.

Keep in mind that you generally may not take third-party materials that you received under another license, and change its license to OpenMDW-1.1 without the licensor’s permission. If you redistribute third-party materials under other licenses, you should be sure to comply with those other licenses’ terms in your distribution.

Yes, but providers of such distributions should be extremely clear and explicit if they do so.

As described above, OpenMDW-1.1 applies not only to machine learning models, but also to other related artifacts that are provided under OpenMDW-1.1.

This means that if other related artifacts are included in the distribution without any other terms, they are presumably considered “Model Materials” and are subject to OpenMDW-1.1’s license grants.

However, a distribution might also include content under other licenses. This could include, for example, pre-existing software from a third party under an open source software license such as MIT or Apache-2.0; or even the model provider’s own related artifacts that are provided under other terms. (This is very similar to how open source software projects under one overall license indicated in the top-level LICENSE file may include components that are under different licenses as well.)

In such a case, the model provider should be very clear in the primary documentation (e.g. the top-level README and/or LICENSE files) about the presence of content that is subject to other license terms. The other applicable license terms should be clearly stated, along with an easy-to-understand explanation about which files and content are subject to the non-OpenMDW terms.

A model provider may want to consider distributing such non-OpenMDW artifacts in a separate distribution or repository if feasible, to avoid confusion to downstream users about which portions are subject to OpenMDW-1.1.

Not at present (as of May 2026). The OpenMDW License Agreement, version 1.1, has just been announced and has not yet been submitted to SPDX for consideration for the SPDX License List. We will update this FAQ if and when it is added to the SPDX License List.

In the meantime, users of SPDX short-form identifiers can use the standard LicenseRef- format that SPDX provides for any licenses that are not on the SPDX License List, in a manner such as the following:

SPDX-License-Identifier: LicenseRef-OpenMDW-1.1

An earlier version of OpenMDW is listed on the SPDX License List with the license identified OpenMDW-1.0.

OpenMDW-1.1 does not impose a particular citation format. Model contributors may include a CITATION.cff file or citation instructions in the README file.

Model cards are short documents accompanying trained machine learning models that provide benchmarked evaluation in a variety of conditions. To reference OpenMDW-1.1, it is recommended to set the License field in the model card metadata and link to the full license text of this repository, with reference to the 1.1/ folder. You can also include a short explanation in the usage or limitations section of the model card if it exists.