Why Open‑Source Licenses Must Evolve for AI: Insights from the GitHub Copilot Lawsuits

The Register published an article arguing that open‑source licenses need to break free from their 1980s constraints and continuously evolve to address artificial intelligence.

The author notes that while free software and open‑source licenses originally targeted code in the 1970s‑80s, today they must transform again to handle AI models.

Although AI was born from open‑source software, the copyright‑based free‑software and open‑source licenses that govern software code are not suitable for AI‑generated large language model (LLM) neural networks and datasets.

Because many LLM programming datasets rely heavily on free‑software and open‑source code, specific measures are required.

For example, last year we reported that open‑source authors obtained lawyer credentials to sue GitHub Copilot.

In November 2022, a class‑action lawsuit filed by Matthew Butterick and Joseph Saveri’s firms alleged that Copilot’s training used open‑source code that infringed the owners’ rights.

The suit claims Microsoft, GitHub and OpenAI created an AI coding assistant that depends on “unprecedented‑scale software piracy.”

Copilot, launched in June 2021, was trained on publicly scraped code repositories, many of which require attribution and retain copyright with the original author. The tool was found to use open‑source code without complying with these terms, prompting the lawsuit for large‑scale copyright violations.

This is why OSI executive director Stefano Maffulli and many other leaders in open‑source and AI are collaborating to combine AI and open‑source licensing in a mutually beneficial way.

As the article points out, such litigation is not limited to large corporations. Sean O'Brien, a Yale law‑school cybersecurity lecturer, warns that a new sub‑industry akin to patent trolls will emerge around AI‑generated works, tracking “your” ChatGPT and Copilot code.

German researcher and politician Felix Reda claims all AI‑generated code belongs to the public domain, while U.S. lawyer Richard Santalesa notes distinct contract and copyright issues.

Santalesa believes companies that produce AI‑generated code will treat the material—including the AI‑generated code—as their own property, just like other intellectual‑property assets.

The issue remains under active discussion, and future legal guidelines are expected to address it.