Asahi Linux Drops MIT for the GPL After LLM Rewrites of Its Unreleased Code

What Asahi Linux announced

In its Linux 7.3 progress report, dated 7 November, the Asahi Linux team said it had made "the difficult decision" to begin transitioning all of its repositories to the GNU General Public License version 2 or later. From now on, the team says, all Asahi code will use strong copyleft licences only.

Asahi had deliberately used the permissive MIT licence for as much of its code as it could. The stated reason was to make it easy for the BSDs and other free operating systems to reuse its Apple Silicon work. The report says that approach is no longer viable.

What the developers say went wrong

The project says it learned of multiple attempts to exploit its permissive licensing, with the aim of replacing the project and monetising Linux on Apple Silicon. In the most extreme cases, the report says, feature code still in development was taken directly from developers' personal forks and either published without attribution or rewritten with LLM help to disguise where it came from.

Asahi names no one in the report. It's FOSS points to recent forks in the same space as context, but notes that Asahi did not identify them as the offenders, so no specific project should be read as accused. It is also worth being precise about the role of AI here. The report describes LLMs as a tool for obscuring the origin of copied code, and separately blames the appetite of frontier AI labs for training data as part of the wider climate. It does not say an AI tool stole the code.

Why WINE is the precedent

The report draws an explicit parallel with WINE, which began under the MIT licence and moved to the LGPL in 2002 after proprietary forks threatened to fragment the project. Asahi's argument is that the same pressure has returned, this time with a tool that makes it cheaper to launder someone else's work.

The new element is rewriting. A copyleft licence obliges anyone who distributes a derivative work to share its source under the same terms, but an LLM rewrite is designed to make derivation harder to see and prove. So the GPL gives Asahi stronger terms without solving detection. Whether a machine rewrite of a codebase counts as a derivative work is still an unsettled question, and Asahi's move does not settle it.

Asahi's stance on AI was already strict

This is not the project's first conflict with generative tools. Its LLM policy broadly forbids generative AI for material contributions, arguing that such systems pose particular legal risks to clean-room reverse engineering because they could taint the result. The policy warns that a contributor found to have concealed extensive LLM use will be banned immediately.

The open question: consent from past contributors

Relicensing an established codebase is rarely simple. OpenBSD developer Bryan Steele criticised the move, saying it appeared to be happening without the consent of all prior contributors, according to It's FOSS. Neither the progress report nor It's FOSS explains how existing contributions will be handled, and the report's own wording is that the team will begin the process.

The cost is real for the projects Asahi once set out to help. BSD systems that relied on permissively licensed Apple Silicon code may not be able to take future Asahi work into their own trees on the same terms. For maintainers elsewhere, the episode is a warning that the bargain behind permissive licensing assumed copying would be visible. Cheap, automated rewriting weakens that assumption, and more projects may decide that copyleft is worth the lost reuse.