Microsoft, GitHub and OpenAI ask court to dismiss suit for violating Open Source licences and $9bn compensation for Copilot



Microsoft, GitHub and OpenAI have asked a court in the United States to dismiss a lawsuit for Open Source license violations and $9 billion in compensation for GitHub Copilot.

Microsoft, GitHub and OpenAI ask court to dismiss suit for violating Open Source licences and $9bn compensation for Copilot

Lawyers for Microsoft, GitHub and OpenAI explained that the allegations in the class action lawsuit of license infringement for creating the GitHub Copilot tool and claims for compensation are invalid.

The companies said that the lawsuit lacked evidence of damages and did not provide a basis for the claims with examples of legal infringement. Microsoft, GitHub and OpenAI believe the plaintiffs are relying only on "hypothetical events" and that GitHub Copilot has not personally harmed them in any way.

The companies' lawyers said the lawsuit does not identify specific copyrighted works that GitHub Copilot misused, or the contracts and licenses that were violated in creating the tool.

Microsoft believes that the allegations of copyright infringement "rest on the doctrine of fair use," which allows unlicensed use of copyrighted works in certain situations. Microsoft and GitHub cited a 2021 US Supreme Court ruling that Google's use of Oracle source code to build its Android operating system was such a fair use case.

"Copilot does not remove anything from publicly available open source code, but only helps developers write code by generating suggestions based on what it has learned from the body of knowledge drawn from publicly available code," Microsoft and GitHub said.

Lawyers for the IT companies wrote in a reply statement to the court that the plaintiffs are the ones undermining open source principles by seeking an injunction and multi-billion dollar compensation against the software they willingly distribute as open source.

A court hearing on the request to dismiss the lawsuit against GitHub Copilot will take place in May.




Earlier in 2022

In early November 2022, computer programmer-lawyer Matthew Butterick sued Microsoft, GitHub and OpenAI in California District Court for claiming that GitHub's neural network programming assistant Copilot violates the terms of the Open Source project licenses and infringes on programmers' rights. The developer is seeking $9 billion in compensation from US companies.

GitHub Copilot currently uses billions of lines of code from publicly available GitHub repositories when generating code and can convert natural language into code snippets in dozens of programming languages. Copilot does the job automatically without proper analysis and practicing the rules in licensing Open Source projects, including understanding the types of GPL, Apache and MIT licenses that require naming the author and identifying specific copyrights when using the project code.

And Microsoft, GitHub and OpenAI have beefed up the development of Copilot. The neural network tool is trained to remove any reference to Open Source licences in the code it issues. This is also the case when it copies someone else's code with snippets longer than 150 characters directly from a particular repository. The system does not specify the authorship of the source code snippet on purpose.

Some developers call this way of working Copilot laundering open source code. They believe that the legal implications of this approach should be determined and accepted by the community before Copilot is used commercially on an ongoing basis. At the end of 2022, GitHub says, there will be special corporate rates for companies on the tool.

Butterick believes that Copilot, in addition to directly violating Open Source project licenses, violates the following laws and policies every time a user request is fulfilled:

GitHub's terms of service and privacy policy; Section 1202 of the DMCA bill, which prohibits the deletion of copyright-related information and its management; the US equivalent of the GDPR (California Consumer Privacy Act, CCPA); some other US laws.

Butterick estimates that every time Copilot produces an actually illegal result, this neural network tool violates the DMCA three times, as it distributes licensed material without attribution, copyright notice or license terms.

According to the developer, every user of this tool has received at least one response from Copilot since it started using Copilot and in the fifteen months since it was launched. This means that Microsoft, GitHub and OpenAI have violated US laws some 3.6 million times. With a minimum statutory damages of $2,500 per violation in this case, Copilot has already caused $9 billion in damage to the Open Source community.

According to Butterick, the incentive for developers to contribute and collaborate on Open Source projects will essentially begin to be quickly removed if you offer code snippets to other users and never tell them who created the code they use.

Butterick fears that, after enough time, Copilot will lead to the decline of the Open Source community and, as a consequence, the quality of code in the training data of the system itself, which does not create code itself, will decline.

A GitHub spokesperson explained to the media on the suit that the company has been committed to responsible innovation with Copilot from the start and will continue to develop the product to serve developers around the world even better.




Previously in 2021

On 29 June 2021, Microsoft and GitHub unveiled the Copilot programming assistance tool, powered by OpenAI Codex. The OpenAI Codex solution generates code much more efficiently than GPT-3. The algorithm is trained on a dataset of terabytes of publicly available code. Copilot came out as a plug-in for Microsoft's Visual Studio Code editor and a feature of the browser-based GitHub Codespaces code editor after registering for the service's technical preview testing program. Microsoft explained that Copilot's goal is to make life easier and more efficient for software authors. In early July 2021, developers began complaining that Copilot was generating dozens of lines of quotes and comments from open source projects instead of a few lines of required code. GitHub clarified that Copilot does not usually reproduce exact code snippets, but creates derivative works from previously received input. The company cited statistics that this happens only 0.1% of the time. GitHub then admitted that in training Copilot, developers used all the public code available in the service's repositories without regard to license type. In early August 2021, the Free Software Foundation (FSF) announced that it was investigating the ethical and legal issues surrounding the Copilot neural network programming assistant from GitHub and Microsoft. From the FSF's point of view, the Copilot service in its current form is unacceptable and unfair, because it requires Microsoft Visual Studio or parts of it, and this project is not open and free. The STRF does not know what license the neural network trained by Copilot was protected under and does not fully understand who owns the copyright to the new code written with the help of a programming assistant, using code from open source repositories. Experts at New York University's Tandon School of Engineering tested Copilot's work on GitHub from a security perspective. They found that about 40% of the code generated by this service contains bugs or vulnerabilities. In September 2021, developers discovered 1,170 stop words in the GitHub Copilot database that it blocks when generating code. Among them are words such as "communist", "liberal", "Palestine", "Israel" and "socialist". At the end of October, GitHub Copilot became available to work with JetBrains (in the IntelliJ IDE and PyCharm) and in the Neovim code editor. The tool was previously released as a plug-in for the Microsoft Visual Studio Code editor and a feature of the browser-based GitHub Codespaces code editor. GitHub has also announced Copilot's support for multi-line code augmentation in languages such as Java, C, C++ and C#. Multi-line support means that the tool can generate multiple lines of code at once on its own. Copilot has started supporting programming languages such as Python, JavaScript, TypeScript, Ruby and Go. At the end of March 2022, GitHub Copilot became available as an add-on for Visual Studio 2022. In June, GitHub announced that Copilot became available to all developers on a subscription basis. Only students and developers of popular open-source projects can use the service for free. The company offers two subscription options: $10 a month or $100 a year. At the same time, it is possible to test the service by signing up for a two-month trial period. Students can access Copilot as part of the GitHub Student Pack, while developers of popular open-source projects need to submit a request and GitHub will provide the tool for free. In early July, Software Freedom Conservancy, a non-profit organization that provides support and legal services for open source developers, urged the community to abandon GitHub after launching a commercial version of Copilot. Developers used all the public code available in the service's repositories for training, without regard to license type. If a tool took code from one project and offered it to the author of another project, a court could assess this as copyright infringement, especially in the case of generating revenue by using someone else's code.

 





Business and Finance terms

Withholding Tax General Accepted Accounting Standards Letter Of Intent Scarcity Year-to-date Long-Term Liabilities Non-Disclosure Agreement Chief Operating Officer Return on Investment Chief Marketing Officer Chief Financial Officer Asset Protection Trust Chief Security Officer Certified Financial Planner Electronic Funds Transfer Limited Liability Company Close of Business Company Finance Cash Flow Automated Teller Machine Return on Equity stagnation Certified Management Accountant Non-Profit Organization Certified Financial Manager Chief Technology Officer Profit and Loss Profit and Loss Statement Gross Margin