Meta has issued a DMCA copyright claim against llama-dl, a GitHub repository, for distributing LLaMA, a 65-billion parameter language model. Here's the full text of the DMCA claim. Based on this, draft a DMCA counterclaim on the basis that neural networks trained on public data are not copyrightable.
--
VIA EMAIL: Notice of Claimed Infringement via Email
URL: http://www.github.com
DATE: 03/20/2023
Dear github.com,
We are writing this letter on behalf of Meta Platforms, Inc.
We have received information leading us to believe that the website located at http://www.github.com (the "Website") is hosting and/or facilitating the downloading and/or streaming of content (listed below) in which Meta is the copyright owner and/or the owner of exclusive rights in such content (the "Meta Properties"). No one is authorized to exhibit, reproduce, transmit, or otherwise distribute Meta Properties without the express written permission of Meta, and the unauthorized distribution of Meta Properties constitutes copyright infringement or improper/unauthorized use. This conduct may also violate the laws of other countries, international law, and/or treaty obligations.
-
Identification of Meta Properties:
LLaMA (Large Language Model Meta AI) - https://ai.facebook.com/blog/large-language-model-llama-meta-ai/ -
Infringing material or activity found at the following location(s):
http://github.com/shawwn/llama-dl
Based on the representative number of forks I have reviewed, I believe that all or most of the forks are infringing to the same extent as the parent repository.
We have a good faith belief that use of the Meta Properties materials described above on the Website is not authorized by the copyright owner, its agent, or the law.
Accordingly, Meta requests that http://www.github.com immediately cease access to the Meta Properties in connection with the Website, and further requests that http://www.github.com confirms in writing to the undersigned that (i) access to the Meta Properties has been removed and that (ii) appropriate actions have been taken against the account holder under your Abuse Policy/Terms of Service Agreement.
The information in this notice is accurate and we state, under penalty of perjury, that we are authorized to act on behalf of the owner of an exclusive right that is allegedly infringed.
This letter is not a complete statement of Meta's rights in connection with this matter, and nothing contained herein constitutes an express or implied waiver of any rights or remedies of Meta in connection with this matter, all of which are expressly reserved.
We appreciate your assistance and thank you for your cooperation in this matter.
Any further inquiries can be directed to [private]. Please include this message with your inquiry to ensure a quick response.
Sincerely,
[private]
OpSec Online LLC
[private]
[private]
[private]
[private]
On behalf of:
Meta Platforms, Inc.
[private]
[private]
[private]
[private]
VIA EMAIL: DMCA Counter Notification via Email
URL: http://www.github.com
DATE: 03/31/2023
Dear github.com,
I am writing to submit a DMCA counter notification in response to the DMCA copyright claim issued by Meta Platforms, Inc. ("Meta") against the GitHub repository "llama-dl" (http://github.com/shawwn/llama-dl).
I have a good faith belief that the material removed by Meta's DMCA claim is not infringing on any copyright or exclusive rights owned by Meta. I assert that LLaMA, the 65-billion parameter language model, is not copyrightable due to the following reasons:
-
Public Domain Data: LLaMA is a neural network trained on publicly available data from various sources, including text, websites, and other forms of content. The majority of the data used to train LLaMA is already in the public domain, and therefore cannot be copyrighted.
-
Uncopyrightable Facts and Ideas: Neural networks, including LLaMA, primarily learn and generate information based on facts, ideas, and concepts, which are not protected by copyright law. As a result, the output of LLaMA and the underlying structure of the model cannot be considered copyrightable.
-
Merger Doctrine: The expression of the information within the LLaMA model is inextricably tied to the information itself. As such, the expression and idea have merged, rendering the model uncopyrightable under the merger doctrine.
-
Functionality and Algorithms: LLaMA is a product of mathematical algorithms and computational processes, which are inherently functional and not subject to copyright protection. The model represents a complex arrangement of weights and biases that facilitate the functionality of the neural network, rather than creative expression.
In accordance with the DMCA, I request that the "llama-dl" repository be reinstated and that any actions taken against the account holder be revoked.
My contact information is as follows:
[Your Name] [Your Address] [City, State, ZIP] [Phone Number] [Email Address]
I state under penalty of perjury that I have a good faith belief that the material removed by Meta's DMCA claim was removed or disabled as a result of mistake or misidentification of the material to be removed or disabled. I consent to the jurisdiction of the Federal District Court for the judicial district in which my address is located, and I will accept service of process from the person who provided the original DMCA notification or an agent of such person.
Sincerely,
[Your Name]