The Software Freedom Conservancy (SFC), a rights organization providing legal support to free projects and advocating for the compliance with the GPL license, announced the cessation of any use of the GitHub code collaboration platform and urged developers of other open projects to follow their example. The organization also launched an initiative aimed at simplifying the migration of projects from GitHub to more open alternatives, such as CodeBerg (based on Gitea) and SourceHut, or implementing their own development services on servers based on open platforms like Gitea or GitLab Community Edition.
The initiative was prompted by SFC's discontent with GitHub and Microsoft not addressing the ethical and legal nuances of using the source code of free software as a basis for building a machine learning model in the commercial service GitHub Copilot. Representatives of SFC sought to understand whether the created machine learning model is subject to copyright and, if so, to whom those rights belong and how they relate to the rights to the code on which the model is built. It is also currently unclear whether a block of code generated in GitHub Copilot, which replicates code from the projects used to build the model, can be considered a derivative work, and whether including such blocks in proprietary software constitutes a violation of copyleft licenses.
Questions were posed to representatives of Microsoft and GitHub regarding the legal norms that underlie the statements made by GitHub's director that training the machine learning model on publicly available data falls under the category of fair use, and that the processing of code in GitHub Copilot can be likened to the use of a compiler. Additionally, Microsoft was asked to provide a list of licenses and a catalog of repository names used in training the model.
A question was also raised about how to reconcile the statement that it is permissible to train a model on any code without regard for the licenses used, with the fact that only open source code was used to train GitHub Copilot and that training does not cover code from closed repositories and proprietary products of the company, such as Windows and MS Office. If training a model on any code is considered fair use, then why does Microsoft value its intellectual property more than that of open source developers?
Microsoft has evaded questions and has not provided a legal analysis confirming the legitimacy of its claims regarding fair use. Attempts to obtain the necessary information began in July of last year. Initially, representatives from Microsoft and GitHub promised to respond shortly but never did. Six months later, a public discussion was initiated regarding potential legal and ethical issues in machine learning systems, but Microsoft representatives ignored the invitation to participate. Ultimately, a year later, Microsoft representatives directly refused to discuss this issue, stating that the discussion is pointless, as it is unlikely to change the position of SFC.
In addition to the claims related to the GitHub Copilot project, the following issues with GitHub are also noted:
- GitHub has entered into a contract to provide commercial services to the U.S. Immigration and Customs Enforcement (ICE), whose activities are viewed by activists as unethical, for example, due to the practice of separating children from their parents after detaining undocumented immigrants. Attempts to discuss the issue of GitHub's cooperation with ICE have encountered dismissive and hypocritical attitudes toward the matter.
- GitHub assures the community of its support for open source software, but the site and the entire GitHub service are proprietary, with a closed codebase that is not available for analysis. Although the Git toolkit was developed to replace the proprietary BitKeeper and aimed to move away from centralization to a distributed development model, GitHub, by providing specific extensions over Git, ties developers to a centralized proprietary site controlled by a single commercial company.
- GitHub executives criticize copyleft and GPL, advocating for permissive licenses. GitHub is owned by Microsoft, which has a history of attacks on open-source software and actions against the copyleft licensing model.
Additionally, it is noted that the Software Freedom Conservancy (SFC) has suspended the acceptance of new projects that do not plan to migrate away from GitHub. For projects already part of the SFC, leaving GitHub is not mandated, but the organization is ready to provide all necessary resources and support should they choose to transition to another platform. Besides its advocacy work, the SFC also collects sponsorship funds and provides legal protection for free projects, taking on the responsibilities of fundraising and managing project assets, thus relieving developers of personal liability in the event of legal disputes. Notable projects supported by the SFC include Git, CoreBoot, Wine, Samba, OpenWrt, QEMU, Mercurial, BusyBox, Inkscape, and about a dozen other free projects.
Source: opennet.ru
