AI can rewrite open source code—but can it rewrite the license, too?

In the realm of software development, engineers have traditionally utilized reverse engineering to replicate the functions of a program without directly copying its copyright-protected code. However, recent advancements in AI coding tools are complicating the legal, ethical, and practical aspects of this 'clean room' rewrite process. A significant controversy erupted last week following the release of version 7.0 of chardet, a well-known open-source Python library designed for automatic character encoding detection. Originally created by Mark Pilgrim in 2006 and governed by a strict LGPL license, chardet has seen various updates over the years. Dan Blanchard has been the library's maintainer since 2012 and recently revamped chardet, describing the latest version as a complete rework under the more permissive MIT license. He collaborated with an AI tool named Claude Code to enhance the library's performance, achieving a remarkable 48-fold increase in speed. Blanchard expressed his long-standing ambition to integrate chardet into the Python standard library but felt hindered by issues related to its licensing and performance. However, his overhaul has not been met with universal approval. A user named Mark Pilgrim resurfaced on GitHub, contending that the new release constitutes an improper re-licensing of his original code, which must remain under the LGPL license. Pilgrim's argument emphasizes that any modifications to his original work should preserve the same licensing terms, raising important questions about how AI-generated code is governed under current copyright laws.