What Happened
In October 2022, a class-action lawsuit was filed against GitHub, Microsoft, and OpenAI alleging that GitHub Copilot violated open-source software licenses by reproducing code from public repositories without providing required attribution or license notices. The suit argued that Copilot was trained on code from public repositories including GPL-licensed code that requires derivative works to maintain the same license and attribution.
Researchers demonstrated that Copilot could reproduce substantial code snippets — sometimes entire functions — that were identical to code in its training data, complete with original comments. When these snippets came from GPL or similar copyleft-licensed repositories, the reproduction without attribution potentially violated those licenses.
Timeline
GitHub Copilot launched in June 2022. Within months, developers documented instances of the tool reproducing recognizable code from open-source projects. In October 2022, Matthew Butterick and the Joseph Saveri Law Firm filed the class-action suit. The case proceeded through preliminary proceedings in 2023-2024, with the court narrowing some claims while allowing the core allegations to proceed.
Impact
The case raised fundamental legal questions about the intersection of AI training, code generation, and open-source licensing. If AI models trained on GPL code must comply with GPL terms, the implications would be enormous for every AI coding assistant in existence. The case potentially affected the legality of training AI on any copyleft-licensed material.
The lawsuit also influenced how AI coding tools were developed and marketed going forward, with companies adding attribution features and adjusting their training data approaches to reduce legal exposure.
Response
GitHub added a code attribution feature that could identify when suggestions closely matched known public code, allowing users to review the source and its license. The company maintained that Copilot’s outputs were transformative and did not constitute derivative works under open-source licenses.
Microsoft argued that AI-generated code completions are analogous to autocomplete features and do not carry license obligations from training data.
Lessons Learned
The Copilot litigation highlighted the unresolved tension between AI training practices and intellectual property frameworks designed for a pre-AI era. Open-source licenses were written assuming human developers would copy and modify code, not that AI systems would ingest millions of repositories and produce outputs that might reproduce licensed content without attribution.
The case demonstrated that the legal frameworks governing AI outputs remain underdeveloped and that companies deploying AI tools face significant legal uncertainty when their models are trained on content with specific licensing requirements.