[License-discuss] Pre-submission discussion: Model Chain Attribution License (MCAL)
Михаил Чернов
podumatos.dev at gmail.com
Fri Oct 16 12:52:42 UTC 2026
Hello,
I would like to get feedback on a new license before any formal submission.
This is about the Model Chain Attribution License (MCAL), which is not a
modification of any existing license but a new license designed for a
specific, measurable problem in the AI supply chain.
The problem is what a 2026 study calls "permissive washing." The study
audited 294,012 dataset -> model -> application supply chains and found
that 96.9% of datasets and 96.4% of models lack the required license text,
even when labeled MIT or Apache-2.0. More importantly, attribution
collapses as artifacts move downstream: only 25.0% of models preserve
compliant dataset notices, and only 5.4% of applications preserve compliant
model notices.
This means that even when upstream components have permissive licenses, the
legal basis for using them is often lost by the time they reach an
application. Companies using these models may be operating outside the
scope of the original license without knowing it.
The MCAL proposes a different mechanism to address this. Instead of
requiring a license file or a copyright notice in every artifact, it
requires a structured, machine-readable Attribution Record. This is a
single, extensible JSON, YAML, or XML file that lists the artifact's
upstream dependencies, their licenses, and their relationship types (e.g.,
"trained-on", "derived-from"). The record propagates automatically through
the chain: from dataset to model to application.
The core question is: is this a real problem that a license can address?
And if so, is a machine-readable attribution record the right mechanism?
I have attached the full text for context. I would appreciate any feedback,
especially on:
1. Whether this problem is worth solving at the license level.
2. Whether the proposed mechanism is feasible and clear.
3. Whether there are any OSD concerns I should be aware of.
Thank you for your time.
--
Mikhail Chernov
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensource.org/pipermail/license-discuss_lists.opensource.org/attachments/20261016/230d336a/attachment-0001.htm>
-------------- next part --------------
Model Chain Attribution License (MCAL)
Version 1.0
Preamble
This License governs the use, reproduction, modification, and distribution of the accompanying artificial intelligence artifacts, including but not limited to model weights, model architectures, training and evaluation data, source code, documentation, metadata, and any other associated materials (collectively, the "Licensed AI Materials"). By exercising any rights granted hereunder, you accept and agree to be bound by the terms and conditions of this License.
1. Definitions
1.1. "License" means this Model Chain Attribution License, Version 1.0.
1.2. "Licensor" means the copyright holder(s), patent holder(s), database right holder(s), or other rights holder(s) who have applied this License to the Licensed AI Materials.
1.3. "You" (or "Your") means an individual or a legal entity exercising rights under this License.
1.4. "Licensed AI Materials" means the materials provided to you under this License, consisting of:
(a) one or more artificial intelligence models, including architecture definitions, learned parameters, weights, and configuration files;
(b) all related artifacts, including training data, evaluation data, source code for preprocessing, training, and inference, documentation, metadata, and supporting tools;
(c) any Derivative Works of the foregoing.
1.5. "Derivative Work" means any work that is derived from, based on, or incorporates the Licensed AI Materials, including but not limited to:
(a) a modified version of a model, such as one created through fine-tuning, domain-adaptive training, quantization, or distillation;
(b) a new model, dataset, or software artifact that incorporates or is trained on the Licensed AI Materials;
(c) any work that includes a substantial portion of the Licensed AI Materials and is identified as a Derivative Work in an Attribution Record.
For the avoidance of doubt, a work shall be considered a Derivative Work for the purposes of this License if it is listed as "derived-from" or "trained-on" the Licensed AI Materials in an Attribution Record.
1.6. "Output" means any content, data, or other material generated by using the Licensed AI Materials, including but not limited to text, images, audio, code, or predictions.
1.7. "Attribution Record" means a structured, machine-readable file
or metadata record that identifies all upstream AI artifacts, their
respective licenses, copyright notices, and origin information, in
a format that can be automatically processed by software tools. The
preferred formats are JSON, YAML, or XML.
An Attribution Record must include, at minimum, the following fields:
(a) For the artifact itself:
(i) artifact identifier;
(ii) artifact type;
(iii) license identifier (SPDX or equivalent);
(iv) copyright holder;
(v) source URL.
(b) For each upstream artifact listed in the Attribution Record:
(i) artifact identifier;
(ii) artifact type;
(iii) license identifier (SPDX or equivalent);
(iv) copyright holder;
(v) source URL;
(vi) relationship type.
Permissible relationship types include, without limitation,
"trained-on", "derived-from", and "distributed-with". The structure
may be extended with additional fields, provided that all required
fields are present and machine-readable.
A non-normative example of an Attribution Record in JSON format is
provided below. The example is illustrative only; the normative
requirements are set forth in Section 4.2.
{
"artifact": {
"id": "example-corp/Example-Model-7B",
"type": "model",
"version": "1.0",
"url": "https://huggingface.co/example-corp/Example-Model-7B"
},
"license": {
"spdx": "MCAL-1.0",
"url": "https://example.org/licenses/MCAL-1.0.txt"
},
"copyright_holder": "Example Corp",
"upstream": [
{
"id": "example-data/Example-Corpus",
"type": "dataset",
"license": "MIT",
"copyright_holder": "Example Data Org",
"relationship": "trained-on",
"url": "https://huggingface.co/datasets/example-data/Example-Corpus"
},
{
"id": "open-foundry/Example-Base-1B",
"type": "model",
"license": "Apache-2.0",
"copyright_holder": "Open Foundry Collective",
"relationship": "derived-from",
"url": "https://huggingface.co/open-foundry/Example-Base-1B"
}
]
}
1.8. "Distribute" means to make available to any third party, whether by sale, gift, license, or any other means, the Licensed AI Materials or any Derivative Work.
2. Grant of Copyright, Patent, and Database Rights
Subject to the terms and conditions of this License, the Licensor hereby grants to You a worldwide, non-exclusive, royalty-free, perpetual, irrevocable license to:
(a) use, reproduce, distribute, and publicly display the Licensed AI Materials;
(b) prepare Derivative Works of the Licensed AI Materials;
(c) modify, merge, publish, and sublicense the Licensed AI Materials and Derivative Works;
(d) extract, reuse, reproduce, and share all or a substantial portion of the contents of any data or database included in the Licensed AI Materials;
(e) use the Licensed AI Materials and Derivative Works for any lawful purpose, including commercial purposes.
This grant includes all copyright, patent, and database rights included or embodied in the Licensed AI Materials. For the avoidance of doubt, this License does not grant any rights in trade secrets, as the Licensed AI Materials are distributed openly and any trade secret protection therein is extinguished by such distribution.
3. Grant of Patent License
Subject to the terms and conditions of this License, the Licensor hereby grants to You a worldwide, non-exclusive, royalty-free, perpetual, irrevocable patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Licensed AI Materials, where such license applies only to those patent claims licensable by the Licensor that are necessarily infringed by the Licensed AI Materials alone or by combination of the Licensed AI Materials with the work to which such Licensed AI Materials was submitted.
3.1. Patent Termination. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Licensed AI Materials or a Derivative Work directly or indirectly infringes any patent, then any patent licenses granted to You under this License for the Licensed AI Materials shall terminate as of the date such litigation is filed.
3.2. No Termination for Copyright Claims. For the avoidance of doubt, nothing in this License shall be construed to terminate any rights granted hereunder as a result of any claim, lawsuit, or proceeding relating to copyright infringement.
4. Attribution Record Requirement
4.1. Any Distribution of the Licensed AI Materials or any Derivative Work must be accompanied by an Attribution Record.
4.2. The Attribution Record must include the fields specified in
Section 1.7 for the Licensed AI Materials and for each upstream
artifact. The Attribution Record must also include the source URL
or repository where the Licensed AI Materials were obtained, if
not already included.
4.3. If You use the Licensed AI Materials to create a Derivative Work, You must include the original Attribution Record, supplemented with information about Your Derivative Work, in any Distribution of Your Derivative Work.
4.4. The Attribution Record must be provided in a machine-readable format (e.g., JSON, YAML, XML) and must be made available at the point of distribution, in a manner that is reasonably accessible to recipients of the Licensed AI Materials or any Derivative Work. Where the Licensed AI Materials or a Derivative Work are distributed via a repository, model hub, or package, the Attribution Record shall be included in the same location as the primary distribution artifact. Where distribution occurs through an API or embedded system, the Attribution Record shall be made available in documentation or metadata associated with the distribution.
4.5. You may not remove, alter, or obscure any copyright notice, license notice, or Attribution Record contained in the Licensed AI Materials.
5. Output Clause
This License does not impose any restrictions or obligations with respect to any use, modification, or sharing of any Output generated by using the Licensed AI Materials. The Licensor claims no rights in Output.
6. Conditions
6.1. License Retention. Any Distribution of the Licensed AI Materials or any Derivative Work must include a copy of this License.
6.2. Notice Retention. You must retain all copyright, patent, and database notices present in the Licensed AI Materials.
6.3. No Endorsement. This License does not grant permission to use the trade names, trademarks, service marks, or product names of the Licensor, except as required for reasonable and customary use in describing the origin of the Licensed AI Materials.
7. Disclaimer of Warranty
THE LICENSED AI MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, AND NON-INFRINGEMENT. THE LICENSOR MAKES NO WARRANTY THAT THE LICENSED AI MATERIALS WILL MEET YOUR REQUIREMENTS OR THAT THEIR OPERATION WILL BE UNINTERRUPTED OR ERROR-FREE. YOU ASSUME ALL RISKS ASSOCIATED WITH THE USE OF THE LICENSED AI MATERIALS.
8. Limitation of Liability
IN NO EVENT SHALL THE LICENSOR BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING FROM, OUT OF, OR IN CONNECTION WITH THE LICENSED AI MATERIALS OR THE USE OR OTHER DEALINGS IN THE LICENSED AI MATERIALS, INCLUDING BUT NOT LIMITED TO DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES, EVEN IF THE LICENSOR HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
9. Termination
9.1. Termination for Breach. If You breach any term of this License (other than Section 3.1), Your rights under this License shall not terminate automatically. The Licensor shall provide You with written notice of the breach, and You shall have thirty (30) days from receipt of such notice to cure the breach. If You fail to cure the breach within such period, this License and all rights granted to You hereunder shall terminate.
9.2. Termination for Patent Litigation. If You institute patent litigation as described in Section 3.1, the patent licenses granted to You under this License shall terminate immediately as of the date such litigation is filed. All other rights granted to You under this License shall remain in full force and effect.
9.3. Survival. Sections 4, 5, 7, 8, and 10 shall survive any
termination of this License.
10. Miscellaneous
10.1. Severability. If any provision of this License is held to be unenforceable or invalid, such provision shall be reformed only to the extent necessary to make it enforceable, and the remaining provisions shall remain in full force and effect.
10.2. No Waiver. The failure of the Licensor to enforce any right or provision of this License shall not constitute a waiver of such right or provision.
10.3. Entire Agreement. This License constitutes the entire agreement between the parties with respect to the Licensed AI Materials and supersedes all prior or contemporaneous understandings and agreements, whether written or oral.
End of License Terms
More information about the License-discuss
mailing list