Anti-AI open source projects gain attention after Torvalds' Linux stance
Linus Torvalds said Linux is not anti-AI, while Codeberg and BSD projects show how divided the software world is becoming.
By Renata Fuchs · Policy Reporter
· 3 min read
Anti AI open source projects are drawing new scrutiny after Linus Torvalds said Linux is not an anti-AI project, a pragmatic position that has exposed a split inside free and open source software over LLM-generated code, funding and governance. The immediate industry issue is not a new product launch or financing round, and no money was disclosed, but the debate affects the code supply chain that startups, cloud vendors and enterprise software teams depend on.
European volunteer-run code forge Codeberg has moved beyond rhetoric. It changed its Terms of Use to bar projects that “mostly consist of code written by ‘generative AI’ tools,” and it also bans cryptocurrency-related projects. Codeberg has separately said it is facing hate campaigns from “far-right forces,” linking the AI debate with a broader fight over the politics of open source infrastructure.
What does anti-AI open source mean?
In this context, anti-AI open source means projects or communities that reject LLM-generated code, built-in AI features or AI-assisted development workflows. The objections vary: copyright status, training data provenance, environmental costs, effects on workers and whether maintainers can review machine-produced contributions at the quality level they require.
The BSD world has already put firmer lines around the issue than Linux. NetBSD has treated LLM-generated code as presumptively tainted since 2024. OpenBSD project lead Theo de Raadt has said LLM-generated code cannot be contributed because it cannot be copyrighted, although OpenBSD still includes outside projects such as tmux that accept LLM-assisted work. FreeBSD was still weighing its policy a year after NetBSD set its guidelines.
Linux distributions are likely to face the same pressure, especially because large FOSS projects such as systemd, GNOME, Wayland and Flatpak have received major development support from Red Hat, now an IBM subsidiary. Red Hat management has promoted AI, according to prior reporting by The Register, but the company has not disclosed a policy that would satisfy maintainers who want AI-free dependency chains.
Browsers are becoming the next fight
Mozilla’s Firefox has become a flashpoint because Mozilla has continued adding LLM-based functions, while giving users options to disable them rather than shipping a separate AI-free edition. Sarah Jamie Lewis, executive director of the Open Privacy Research Society, has called for cooperation among Firefox forks and estimated that about 136,000 lines of Firefox code exist only to support built-in LLM, chat or local AI features.
Lewis has created a Base Browser Project repository and a preliminary list of features to remove. The task is difficult because viable browser engines are scarce. Servo is being revived but remains incomplete, while Ladybird has recently adopted AI coding. Existing Firefox forks such as Waterfox and Librewolf are relevant because they already remove telemetry or privacy-sensitive Mozilla components. Waterfox describes itself as AI-free, and Librewolf says it is focused on “privacy, security, and freedom.”
Those alliances are not straightforward. Tor Browser and Mullvad Browser also remove significant Mozilla code, but Mullvad has drawn criticism after reports that a co-owner donated millions to an anti-immigration populist political party. Mullvad accepts several payment methods, including Bitcoin, Bitcoin Cash and Monero, according to its pricing page, putting privacy, VPNs and cryptocurrency into the same operational discussion.
The same tension appears around projects adjacent to anti-AI software. Xlibre, an X11 server used by Vendefoul Wolf Linux, is described as explicitly anti-DEI, and members have publicly made comments critical of transgender people. Similar controversies have touched Hyprland and Ruby on Rails creator David Heinemeier Hansson, according to The Register’s prior coverage.
The anti-AI software movement has a shared opponent in LLM-generated and LLM-assisted code, but shared opposition is not the same as shared values. For founders and engineering leaders, the practical takeaway is that “AI-free” may soon become a software procurement claim that requires due diligence on maintainers, dependencies and governance, not just a checkbox in a README.
This story draws on original reporting from The Register.
