Developers to sue Microsoft over training AI with their code.

Independent developers plan to file a class-action suit against Microsoft over the use of their code to train Copilot, an AI assistant. The story is reported by Motherboard.

The initiative is led by programmer and lawyer Matthew Butterick. In his words, the company wants to turn independent developers into “a source of resources that must be mined.”

“Even cows get food and blood from the deal. Copilot does nothing for our individual projects. And nothing for open-source more broadly,” Butterick said.

Earlier, some programmers noticed that the tool partially copies their work. Twitter users documented examples of software—fragments of code that resemble those in their own repositories.

In response to plagiarism accusations, GitHub stated that the tool does not reproduce strings from training data. About 99% of Copilot’s output code does not correspond to the dataset, according to an internal company investigation.

According to GitHub’s license agreement, developers may freely use open-source software provided they credit the original source. Experts say this is virtually impossible in Copilot’s case.

“How can users comply with the license terms if they are not even aware of their existence?” Butterick said.

The lawyer urged independent programmers to contact his team as part of the preparation for a potential class-action suit against the corporation.

“The legality of Copilot should be scrutinised before damage to open-source software becomes irreparable,” Butterick said.

Microsoft did not comment on the potential suit.

In June 2021, GitHub introduced AI-powered code auto-completion tool Copilot, built on the Codex language model from OpenAI.

In July, Copilot was suspected of copying copyrighted fragments of open-source code.

In October, GitHub said the AI assistant helped write 30% of all new code added to repositories.

Subscribe to ForkLog news on Telegram: ForkLog AI — all the news from the world of AI!