import streamlit as st import base64 PDF_WIDTH = 700 PDF_HEIGHT = 1000 PDF_PATH = "/AI%20ORUS%20License%20(1).pdf" def display_pdf(file): # Opening file from file path with open(file, "rb") as f: base64_pdf = base64.b64encode(f.read()).decode('utf-8') # Embedding PDF in HTML pdf_display = F'' # Displaying File st.markdown(pdf_display, unsafe_allow_html=True) st.title("FAQ of the AI Open Responsible Use License") st.markdown("Licensing discussions in the field of AI have been ongoing for quite a while. Even though some proposals were made some time ago on the data licensing front (see [Linux Foundation CDLA](https://www.linuxfoundation.org/press/press-release/enabling-easier-collaboration-on-open-data-for-ai-and-ml-with-cdla-permissive-2-0), [Montreal Data Initiative](https://arxiv.org/abs/1903.12262), [Open Data Commons](https://opendatacommons.org/licenses/)), there has not been major adoption for the moment, compared to Open Source and Creative Commons licenses. However, these licenses are not tailored to ML, and that can cause confusion. When it comes to machine learning (ML) models, 2022 has been the year of new customized ML model licenses such as OPT175, BB3, SEER, LLaMA big models' [licenses](https://github.com/facebookresearch/metaseq/blob/main/projects/OPT/MODEL_LICENSE.md); the GLM130 ML model [licenses](https://github.com/THUDM/GLM-130B/blob/main/MODEL_LICENSE); or [responsible AI licenses](https://www.licenses.ai/blog/2022/8/26/bigscience-open-rail-m-license) (RAILs). However, until now, there was no AI-specific license enabling the licensing of every ML component in a single license with a simple approach similar to the one taken by an MIT license. Today, we present the [AI Open Responsible Use License v.1.0.0](https://huggingface.co/spaces/CarlosMF/AI-ORUS-License-v1.0.0) (AI ORUS License).") st.markdown("## What’s the goal of the license?") st.markdown("The AI ORUS license is an omnibus license that can be used for all parts of ML, not just software, or data, or models alone. It enables maximum freedom -- no copyleft, no notices, no make-work attributions. Finally, the license addresses ethical challenges of potential misuse of ML material with a license limitation and an indemnity clause in case the licensed material is used in a manner not compliant with the law.") st.markdown("## Can I use this license to distribute different materials all in one?") st.markdown("Yes, this license was created to make it easier to put all your ML materials under one license: code, documentation, data, models, or other works. It is a permissive license, so the rules are not complicated or specific to the type of material you provide.") st.markdown("## Is this an Open Source or Creative Commons license?") st.markdown("No, this is a different license for ML-specific purposes. However, the fact that the license is not an OSI-approved or CC license does not mean that the license is not an “open” one. This license has a limitation that requires use of the material to be in compliance with law. Open Source licenses don’t require this. This restriction was included to ensure that the materials being provided will not be used for unlawful purposes.") st.markdown("## What do you mean by “contributor”?") st.markdown ("The contributor is the rights holder who authorizes use under this license of the AI material. That usually means the creator or inventor of the material.") st.markdown("## What do you mean by “material”?") st.markdown("“Material” means the object of the license, in this case, any AI-related component such as source code, datasets, or ML models. This term is purposely broad and neutral. It is not meant to be limiting. You can license any materials you like under this license.") st.markdown("## What do you mean by compliance with law?") st.markdown("That is determined by the local law that applies to you, your business, or your use of the licensed material. If you don’t understand how to comply with law, you should consult with a lawyer. The AI ORUS does not require anything unique to the license–it leaves that definition to lawmakers.") st.markdown("## Is this license compatible with permissive Open Source licenses?") st.markdown("It depends on the license and what you are doing with the material. Generally, the AI ORUS will not be compatible with any copyleft license, because it has a license restriction. So for example, you should not combine AI ORUS software with GPL software in the same program–you won’t be able to comply with GPL. However, you can probably combine AI ORUS software in the same program with LGPL or MPL software. Those rules come from the other licenses, not AI ORUS. AI ORUS does not control other software you combine with AI ORUS software. The same rules roughly hold for material other than software.") st.markdown("Generally, the AI ORUS will be compatible with permissive open source licenses, like Apache 2.0 or MIT. This means you could combine code under both licenses in the same program. But you will have to comply with the requirements of both licenses. The same will roughly be true for proprietary-licensed code.") st.markdown("You can’t “relicense” AI ORUS-licensed material under other licenses. If you redistribute material under the AI ORUS license, you need to pass along the license requirements.") st.markdown("## What’s the difference between this license and the CC0 1.0 Universal license?") st.markdown("In terms of scope: Contrary to the CC0 license, the ORUS license does not solely apply to copyright and neighboring rights. CC0 license refers to “works” for which there might be copyright protection and/or related rights, such as moral rights or even EU sui generis database rights. In contrast, the ORUS license applies to “any rights”, including patent or trade secrets rights.") st.markdown("In terms of the prerogatives enabled by the licensor: Whereas with the CC0 license, the licensor “waives, abandons or surrenders” their rights, with the ORUS license licensor “authorizes” the use of the material for any purpose that would otherwise infringe the rights of the licensor (i.e. licensor does not relinquish any rights).") st.markdown("## Should I specify what I’m licensing in the license text?") st.markdown("You can. However, please be aware that the license might be harder to identify afterward as it can be deemed a different one or a different version for scanning tools. Those tools can fail to identify licenses if even a word is changed. Plus, you might find it a chore to update the license if you change your materials. It is better if you inform recipients about the specific licensed material outside of the license file. But it’s your choice.") st.markdown("## Why use this license instead of the RAIL license?") st.markdown("Compared to the existing adopted versions of RAILs, the AI ORUS license is a simple and less restrictive license. For instance, the only use restriction included in the AI ORUS license is a general requirement to comply with law.") st.markdown("## Why did you include an indemnity? What does that mean?") st.markdown("An indemnity is a way to be sure that users of licensed materials, rather than the creator or inventor (licensors), bear the burden for what users do with the materials. Because the only restriction on use is so general–compliance with law–users have great latitude to do what they think is lawful and right. But the creators or inventors should not bear legal liability for that choice. Indemnities also exist in open source licenses like the Mozilla Public License and the Eclipse Public License. They are intended to provide comfort to creators or inventors who want to use those licenses to make their works available freely to everyone.") st.markdown("## Do I have to click/sign to accept the license?") st.markdown("No. As many other open licenses, you are accepting the license by using the licensed AI material.") st.markdown("## Disclaimer") st.markdown("No legal advice is provided by the authors with this FAQ or the AI ORUS License. Please contact a lawyer in case you have questions regarding the use of this license for AI material you want to distribute or use under this license.") st.markdown("Authors of the license: Heather Meeker, Joseph Jacks, Carlos Muñoz Ferrandis")