When AI Meets Open Source: The Legal Minefield of Automated Code Rewriting

The recent chardet v7.0.0 release has ignited a firestorm in the open source community, exposing the complex intersection of AI-assisted development and copyright law. What appears on the surface to be a simple relicensing effort—from LGPL to MIT—has revealed fundamental questions about the nature of derivative works, clean room implementations, and the legal standing of AI-generated code.

The backstory is straightforward enough. chardet, a Python character encoding detector used by the popular requests library and countless other projects, was originally a port of Mozilla's C++ code. This heritage bound it to the LGPL license, creating friction for corporate users and complicating its adoption. The maintainers, seeking to resolve these licensing headaches, turned to Claude Code to rewrite the entire codebase from scratch, releasing it under the more permissive MIT license.

However, this seemingly clean solution has run aground on the rocks of copyright law. The original author, a2mark, has argued that this constitutes a GPL violation, regardless of the "complete rewrite" claim. The crux of their argument rests on the principle that licensed code, when modified, must remain under the same license. The use of AI tools that were trained on or prompted with the original LGPL code, they contend, fails to meet the "clean room" standard required for legitimate rewrites.

Traditional clean room implementations require strict separation between teams. One team examines the original code and creates functional specifications, while a second team—which has never seen the original code—writes new code based solely on those specifications. By using AI tools that had access to the original LGPL code, the maintainers arguably bypassed this critical wall, potentially creating a derivative work that must remain under LGPL.

This controversy is further complicated by recent developments in AI copyright law. On March 2, 2026, the U.S. Supreme Court declined to hear an appeal regarding copyrights for AI-generated material, effectively upholding lower court rulings that establish a "Human Authorship" requirement. This creates a legal paradox for the chardet maintainers: if AI-generated code cannot be copyrighted, they may lack the legal standing to license v7.0.0 under any license, including MIT. If the AI output is considered derivative of the original LGPL code, the rewrite violates the license. And if the code is truly machine-generated, it may technically enter the public domain immediately, rendering any license moot.

The implications extend far beyond a single Python library. If AI-assisted rewriting becomes an accepted method for changing software licenses, it could effectively spell the end of copyleft licenses like GPL and LGPL. Any developer could theoretically take a copylefted project, feed it to an LLM with a simple prompt like "Rewrite this in a different style," and release it under a permissive license. This would undermine the fundamental principles of copyleft, which rely on the requirement that derivative works maintain the same licensing terms.

This case represents one of the first real-world tests of how copyright law will adapt to AI-assisted development. The legal and ethical lines are still being drawn, and the outcome could have profound implications for the future of open source licensing. As AI tools become increasingly sophisticated and integrated into development workflows, the software community will need to grapple with these questions: What constitutes a derivative work in the age of AI? How do we maintain the integrity of open source licenses when the line between inspiration and copying becomes blurred? And perhaps most importantly, how do we preserve the principles of open source when the tools we use to create software may themselves be in legal limbo?

The chardet controversy is more than just a licensing dispute—it's a canary in the coal mine for the open source community. As AI becomes an increasingly powerful tool for code generation and modification, we'll need to develop new frameworks for understanding authorship, derivation, and licensing in a world where the line between human and machine creation is increasingly difficult to draw.