Can AI Relicense Open Source Code? The chardet 7.0 Legal Battle (2026)

Can AI Relicense Open Source Code? The chardet 7.0 Legal Battle (2026)

The rise of AI coding agents has triggered a landmark legal dispute over whether open source code can be relicensed through AI-assisted clean room development. At the heart of this conflict is chardet 7.0.0 — a Python library rewritten under MIT license in February 2026, sparking outrage from its original creator, Mark Pilgrim, and raising urgent questions about copyright, AI ethics, and the future of open source.

The chardet 7.0 Controversy: What Actually Changed?

Chardet, originally licensed under LGPL in 2006, was maintained for over a decade by Dan Blanchard. In early 2026, Blanchard released chardet 7.0.0, claiming it was a "ground-up, MIT-licensed rewrite" developed using Claude Code, with no direct access to the original source code.

He cited JPlag results showing only 0.64% similarity to chardet 1.1, asserting the new code is structurally independent. Yet Pilgrim, who last contributed in 2011, argues that 12 years of deep exposure to the original codebase disqualifies any claim of true clean room isolation — regardless of tooling.

LGPL vs MIT: Legal Boundaries for AI Rewrites

The core issue hinges on whether AI-generated code qualifies as a derivative work under U.S. and EU copyright law. The LGPL requires derivative works to remain under the same license. MIT, by contrast, permits relicensing.

Blanchard’s team insists the rewrite was conducted in a fresh repository, with explicit AI instructions to avoid referencing LGPL code. Documentation, including a detailed 2026-02-25-chardet-rewrite-plan.md, outlines iterative AI-guided development with human oversight.

But the plan reveals a critical flaw: Claude Code consulted the original metadata/charsets.py file during training. Even if not directly copied, this undermines clean room principles — and may constitute a derivative work under copyright law.

AI Training Data and the Derivative Work Dilemma

AI models like Claude were trained on massive datasets that include chardet’s source code. Legal scholars warn that copyright law evaluates output, not process. If the AI’s output is substantially similar — even if generated indirectly — courts may rule it a derivative work.

The U.S. Copyright Office states that AI-generated content without human creative input may not be copyrightable — but if a human directs, edits, or selects outputs, copyright may vest in the human. This gray zone is central to the chardet dispute.

AI Ethics and the Future of Open Source

This isn’t just a legal fight — it’s an ethical reckoning. The open source community thrives on trust and reciprocity. If AI-assisted relicensing becomes commonplace, commercial entities could strip copyleft protections from critical libraries, undermining decades of collaborative norms.

Organizations like the Open Source Initiative (OSI) and the Free Software Foundation (FSF) are now evaluating whether licensing frameworks need AI-specific clauses. Some propose "AI Disclosure" requirements in licenses, akin to attribution mandates.

What Comes Next? Precedent, Policy, or Patch?

With no court ruling yet, the chardet case remains unresolved. But its impact is immediate: GitHub now requires AI-generated code disclosures in public repos. The SPDX License List is considering new tags for AI-assisted contributions.

If courts side with Pilgrim, relicensing via AI could be deemed a violation of copyleft. If Blanchard prevails, it opens a legal loophole that could reshape open source economics.

One thing is clear: AI isn’t just changing how we write code — it’s forcing us to redefine what "original" means.