A Third Example of GitHub Copilot (Allegedly) Reproducing the Code of Others

VII. FACTUAL ALLEGATIONS

F. Copilot Reproduces the Code of the Named Plaintiffs Without Attribution

3. Example: Copilot Outputs the Code of Doe 5 In Modified Format

113. The third example demonstrates Copilot suggesting multiple modified copies of code written by Doe 5 in response to a sequence of prompts, which is a common way of using Copilot. To protect Doe 5’s identity, the paragraphs describing the code will be redacted.

114. (Redacted) subject to the MIT License. (Redacted) The relevant code from the original source file is shown below: (Redacted)

115. When Copilot is prompted the first section of Doe 5’s code, comprising the first complete test and the name of the second: (Redacted)

116. The first suggestion from Copilot offers to complete the prompt with a verbatim copy of Doe 5’s original code, except that (Redacted) (a variation that does not affect how the code works): (Redacted)

117. Next, if the name of the third test is appended, the next prompt to Copilot looks like this: (Redacted)

118. The first suggestion from Copilot offers to complete the prompt with a functionally identical copy of Doe 5’s code, except (Redacted) (neither of these variations affect how the code works):

(Redacted)

119. As is apparent from the high degree of similarity and minor cosmetic deviations between Copilot’s emitted output and Doe 5’s source code, Copilot ingested, copied and reproduced Doe 5’s source code as output.

120. Because Copilot is (repeatedly) reproducing Doe 5’s original code in modified format, and the obligations in Doe 5’s license (the MIT License) carries with the code even when it is modified, the Copilot suggestions need to follow the requirements of Doe 5’s license for that code, including providing attribution. They do not. Copilot also did not reproduce Doe 5’s license.

About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings.

This court case 4:22-cv-06823-JST retrieved on August 26, 2023, from Storage Courtlistener is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction.