Goals
What are the goals of ivrit.ai?
ivrit.ai‘s goal is to enable advanced AI technologies to read, write, hear and speak Hebrew fluently. We view it as a strategic imperative to provide Hebrew speakers, and their communities, with access to the best AI technology. High quality data is a fundamental requirement for enabling AI models, and the ivrit.ai project focuses on aggregating, tagging and making available such data under a legal framework that is appropriate for building AI technologies.
License
What license is the ivrit.ai material available under?
ivrit.ai makes available material under a somewhat modified version of the Creative Commons Attribution 4.0 International License (CC-BY). For reasons we explain in more detail below, our license only permits use of the licensed material in certain fields and for certain purposes. Uses outside of those specific fields or outside of those purposes is not permitted. As such, the ivrit.ai license is not a “Creative Commons” license – it is a modified version of that license and should not be confused with the licenses made available by the Creative Commons. The ivrit.ai license is not approved or certified by Creative Commons.
What additional restrictions are imposed on materials made available by ivrit.ai?
ivrit.ai imposes two additional restrictions on the use of material that it makes available. First, the material may be used and distributed only for the purpose of training AI models. As such, if you use the material for other purposes aside from training AI models, such use is not licensed by ivrit.ai and you may be infringing copyrights in the material. Second, the material may not be used to create deepfakes. In other words, ivrit.ai makes available the material only for a limited set of purposes, and not for other purposes. Again, this means that ivrit.ai is not making available the material under a “Creative Commons” license, but rather under a modified version of that license that is not approved by Creative Commons.
Why does ivrit.ai impose such additional restrictions?
ivrit.ai does not make available its own material, but rather collects and makes available material from a variety of other sources. These other sources have generously agreed to make available this material for what these sources see as a good cause – training AI models on Hebrew language source material. However, these sources often have sensitivities regarding the unauthorized use of the material. These sources could be commercial entities – and making available the material without these additional restrictions could allow third parties to compete with the entity’s business using its own material. Such sources could also be non-profit or government entities – and such entities often see themselves as custodians of the material and have sensitivities as to how it is used. In other words, without these additional restrictions, it is likely that iivrit.ai would not be able to make available much of the material that it currently provides on its websites. ivrit.ai is able to collect and make available such material only as a result of such additional restrictions.
Doesn’t imposing such additional restrictions mean that the ivrit.ai material is not really open source?
Indeed. Licenses that have restrictions as to their field of use are not considered “open source” according to the widely-accepted criteria of the Open Source Initiative. See here for more details. Nevertheless, ivrit.ai has chosen to make available the material only pursuant to these additional restrictions for the reasons described above. As noted above, without such additional restrictions, ivrit.ai may not be able to make available this material.
Am I allowed to use the ivrit.ai material for commercial purposes?
The Creative Commons attribution licenses allow for the use of the licensed material for commercial purposes and for profit. See generally here. While ivrit.ai does not make available material under the CC-BY license, but rather under a modified version of that license that is not approved by Creative Commons, the additional restrictions imposed by ivrit.ai do not restrict the use of the material for commercial purposes.
If ivrit.ai imposes additional restrictions, why doesn’t the project draft its own license? Why is the ivrit.ai license based on a modified version of the Creative Commons license?
ivrit.ai wants to ensure that the material which it makes available is widely used. As such, ivrit.ai wants to ensure that it uses a license that is readily and broadly understood, and that will not require an army of lawyers to interpret. ivrit.ai believes that a modified version of the CC-BY licenses satisfies that criteria. While the additional restrictions imposed by ivrit.ai add a modicum of complexity, ivrit.ai believes that the scope of such additional restrictions are clear and can be readily understood.
Does Creative Commons allow ivrit.ai to impose such additional restrictions beyond the Creative Commons license?
ivrit.ai and the relevant copyright holders, as applicable, are the licensor of the material made available by ivrit.ai and can choose the license pursuant to which such materials are made available. Section 7(b) of the CC-BY license expressly contemplates terms that may supplement the original CC-BY license. We acknowledge that the Creative Commons foundation itself does not favor the imposition of such supplemental licensing terms. See here for more details. Nevertheless, without such additional restrictions, ivrit.ai would not be able to make available the material it provides. As such, taking into account all facts and circumstances, ivrit.ai has chosen to make available the material under a modified version of the CC-BY license. Again, it should be clarified that the ivrit.ai license is not approved or made available by Creative Commons – it is a license that has been specially drafted by ivrit.ai for its own special circumstances.