Can you use an LLM to rewrite an open-source library and slap a different license on the result? That's the question at the center of a growing dispute over Chardet, a widely-used Python library for detecting character encoding in text files.
Chardet 7.0 shipped as a complete rewrite, claiming up to 41x performance improvements. But the release also changed the project's license from LGPL (which requires derivative works to stay under the same license) to the more permissive MIT license (which lets anyone do almost anything with the code, including using it in proprietary software).
The Original Author's Objection
Mark Pilgrim, who created Chardet, pushed back publicly on GitHub. His argument is straightforward: the developers had "ample exposure" to the LGPL-licensed code, and using an LLM to produce a rewrite doesn't make it a clean-room implementation. Under the LGPL, modified versions must keep the same license. Pilgrim called the relicensing "an explicit violation of the LGPL" and demanded it be reverted.
The counterargument from the Chardet 7.0 team is that an LLM-generated rewrite constitutes a new work. This is legally untested territory, and the answer probably depends on whether courts treat LLM output as a derivative work when the LLM was prompted with or trained on the original code.
Bigger Than One Library
This matters beyond Chardet because the same playbook could be applied to any open-source project. Feed a copyleft-licensed codebase to an LLM, ask it to rewrite the functionality, publish under a permissive license. If that's legal, the entire copyleft model (GPL, LGPL, AGPL) has a massive loophole.
Linux kernel maintainers have already flagged this exact concern. If AI agents can systematically rewrite GPL kernel code and relicense the output, it undermines decades of licensing infrastructure that keeps Linux and thousands of other projects open.
The GitHub discussion around Chardet 7.0 is heated, with no resolution yet. No court has ruled on whether LLM-mediated code rewrites count as derivative works under copyright law. Until one does, expect more of these disputes as AI coding tools make large-scale rewrites trivially easy to produce.