AI Clean Room Engineering: Automated License Stripping and the Threat to Copyleft Open Source

A company called Malice offers "clean room as a service" — automating the traditional clean room reverse engineering process using AI to strip copyleft license obligations from open-source code. ## How It Works The process follows the established legal clean room pattern: **Robot A** reads the original GPL-licensed code and documentation, then generates a functional specification describing what the code does without including the implementation. **Robot B** implements the specification from scratch, never seeing the original code. The output is functionally identical software with a new license and no legal connection to the original. The service accepts a `package.json` and "liberates" dependencies by replacing copyleft-licensed packages with clean-room reimplementations. ## Legal Foundation **Baker v. Selden (1879):** The US Supreme Court ruled that copyright protects expressions but not ideas. An accounting method described in a book could be legally reimplemented by someone else. **Phoenix Technologies IBM BIOS (1984):** The precedent for clean room reverse engineering. Engineer A studied IBM's BIOS and wrote a specification. Engineer B implemented from the specification alone, never seeing the original. Courts ruled this was legal. This precedent enabled the entire IBM-compatible PC industry. AI clean room engineering applies the same legal framework — the AI simply replaces the human engineers. The process is identical; only the speed and cost have changed. ## Implications for Open Source GPL and other copyleft licenses work through a viral mechanism: any derivative work must carry the same license, ensuring corporate users contribute changes back to the community. AI clean room engineering breaks this mechanism by producing code that is not legally a derivative work. If corporations can trivially strip copyleft obligations from dependencies, the incentive structure that sustains open-source contribution is undermined. Contributors who chose GPL specifically to ensure corporate reciprocity lose their leverage. ## The Creators' Argument Presented as a conference talk called "The Death of Open Source," the creators argue that open-source licenses have always been legally fragile, that copyright law fundamentally protects expression rather than behavior, and that AI merely makes the existing vulnerability instant and cheap. Their stated goal is generating enough attention to force legal reform. ## Possible Outcomes Legal reform updating copyright law to address AI-generated derivatives (unlikely in the near term). New license types explicitly covering AI-generated reimplementations. Quiet corporate adoption with no legal challenge. Or sufficient community backlash to create social pressure against using these tools even where legal.