A couple of weeks ago, I wrote about developers threatening a suit against Microsoft/GitHub/Copilot for copyright violations. The issue was that the AI assistant Copilot was “suggesting” code that was an exact copy of copyrighted code without attribution. That post talked about Matthew Butterick, a lawyer and programmer, who was collecting information from developers who felt their rights had been violated by the application.
At that time, a lawsuit was still a tentative outcome but Butterick and his law firm have moved forward and filed a lawsuit against GitHub, Microsoft, and OpenAI for allegedly violating open source licenses. The defendants, of course, are claiming fair use while the plaintiffs are claiming clear violations of their licenses.
Whatever your feelings about this, it’s almost certainly a harbinger of things to come. The issue is that AI systems are being trained on copyrighted material so of course they output results that are arguably copyright violations. If you think Copilot is definitely producing copyright violations, consider things like Dall-E where you describe what you want a picture to be about and it produces something very much like an existing work. Is that a copyright violation? If you’re not a lawyer, it’s a hard question. One thing for sure, the legal system is not (yet) prepared to deal with the question.