You may have heard how GitHub Copilot helps programmers write more code faster. Copilot is an AI trained on open source code, programmed to generate code suggestions for programmers who want help. As a consequence, Copilot’s output is “somehow” (secret machine learning sauce) based on all the open source input code. However, none of the output of Copilot credits the authors of the open source code that Copilot was trained on. If Copilot’s output was a distribution of its input code, it would have to provide at least the copyright notices of the original authors. Given that it doesn’t do this, the Microsoft lawyers obviously believe that Copilot is not distributing the original open source code.
Now, some open source programmers and their lawyers have taken action and are suing Microsoft and others for violating their rights by not complying with the licenses of the input code. The key question is: Is the GitHub Copilot output a derivative of its input code or something entirely different? The answer to this question will have profound consequences for the software industry and beyond.