22 Open-Source Projects That Ban AI-Generated Code (and How to Check Before You PR)

If you use Claude Code, Copilot or Cursor to send pull requests to open-source projects, there is a new list you should bookmark. A tracker called Open Source Says No (osssaysno.com) now catalogues 22 projects that refuse AI-generated contributions.

The list includes heavyweights such as QEMU, Gentoo Linux, Zig, Servo and immich. It went viral on 11 November 2026, the same week Asahi Linux relicensed its code over alleged AI-assisted copying.

This guide lists every project on the tracker, explains why they say no, shows which big projects still allow AI with rules, and gives you exact steps to check a project’s policy before you open a PR.

What Open Source Says No Is

Open Source Says No is a simple public tracker at osssaysno.com. Each entry names a project, summarises its AI policy, lists what the ban covers (code, docs, issues, art, translations) and links to the policy itself.

Every entry also shows a “Verified” date, when the policy was last checked. The list was refreshed on 11 November 2026, and several entries were verified that day.

The site does not name its operator, but it links to a GitHub repository, jlengrand/oss-banning-ai. Anyone can suggest a project by pasting its GitHub URL into a form, which opens a prefilled issue and runs an automated policy check.

Note: The tracker lists projects that say no. Most open-source projects have no AI policy at all, and many more allow AI with conditions. A missing entry does not mean a project welcomes AI PRs.

All 22 Projects and Their Rules

Here is the full list as shown on osssaysno.com on 11 November 2026. Summaries follow the tracker and each project’s linked policy.

Warning: Some rules go further than you might expect. stb rejects bug reports that AI only reviewed, and recursion-context bans the contributor, not just the PR.

Notice that fast-check is different. It targets automated accounts and agents opening PRs on their own, not a human who used AI while writing.

Why Projects Are Saying No

The policies give four main reasons. Most projects cite more than one.

• Copyright and the DCO. QEMU says the copyright and licence status of AI output is “ill-defined”, and it “is not willing or able to accept the legal risks”. Contributors sign a Developer Certificate of Origin they cannot honestly sign for AI output.
• Reviewer overload. awesome-claude-skills says it stopped AI PRs because of sheer volume. Generating a PR now takes seconds, but reviewing it still takes a human’s time.
• Quality and trust. Fyrox and others reject PRs they merely suspect are AI-made, because confident-looking but wrong code is costly to catch.
• Licence laundering. AI can rewrite code until its origin is hidden, which strips attribution and licence terms. That is the core of the Asahi Linux dispute below.

Key point: QEMU still allows AI for research, static analysis and debugging, as long as the AI’s output is not in the patch. “No AI contributions” rarely means “never open ChatGPT”.

Big Projects That Allow AI, With Rules

The opposite camp is just as important to know. Several major projects accept AI-assisted work if a human stays accountable.

The kernel’s own example trailer is Assisted-by: Claude:claude-3-opus coccinelle sparse. Basic tools such as git, gcc and editors should not be listed.

Tip: The pip and attrs rule clashes with tools that add a co-author line by default. Check your agent’s commit settings before contributing there.

The Asahi Linux Twist

The tracker’s timing is no accident. In its 7.3 progress report (early November 2026), Asahi Linux, the project that runs Linux on Apple Silicon Macs, said it is moving all its repositories from mostly MIT to GPL-2.0-or-later.

According to It’s FOSS, the report says in-development code was taken from personal forks before release, and others republished code without credit or rewrote it with AI to hide its origin. Asahi did not name anyone.

The MIT licence, the report says, “explicitly allows this, and has no carveouts to protect against bad faith actors.” OpenBSD developer Bryan Steele criticised the move, saying it happened “seemingly without the consent of all prior contributors.”

Asahi’s separate LLM policy already “broadly forbids” generative AI for material contributions. An LLM-drafted GitHub issue may simply be closed, but hiding extensive LLM use leads to an immediate ban.

How to Check a Project’s AI Policy

Do this every time before you point an AI agent at someone else’s repository. It takes two minutes.

1. Search the tracker. Open osssaysno.com and look for the project name.
2. Read CONTRIBUTING.md. Most policies live here, often under a heading like “AI” or “LLM usage”. Some projects use a separate AI_POLICY.md (gh-dash, fast-check) or a docs page (QEMU, Servo, Gentoo).
3. Grep for keywords. On the command line, fetch the file and search it. See the command below.
4. Check the PR template. Look in .github/pull_request_template.md for an AI disclosure checkbox.
5. Search closed PRs and issues. On GitHub, search the repo for “AI-generated” or “LLM” to see how maintainers actually respond.
6. Ask if unsure. Open a short discussion or mailing-list post before writing code. QEMU explicitly welcomes exception requests on qemu-devel.

Quick keyword check (replace OWNER and REPO):

curl -s https://raw.githubusercontent.com/OWNER/REPO/HEAD/CONTRIBUTING.md | grep -inE "llm|generative|ai-generated|copilot|chatgpt|claude"

Empty output does not prove a project allows AI. The policy may live in another file, a wiki page or a mailing-list decision.

Tip: Put the result in your agent’s instructions. Our AGENTS.md guide shows how to tell coding agents a repo’s rules, such as “never open PRs here”.

How to Disclose AI Help Where It Is Allowed

Where AI is allowed, disclosure is usually expected or encouraged. Follow the project’s exact format if it has one.

• Commit trailer: Git 2.32+ can add trailers for you: git commit -s --trailer "Assisted-by: Claude:claude-opus-5-5". The -s adds your own human sign-off.
• PR description: say briefly what the AI did, such as “drafted tests, I reviewed and edited them”.
• Review everything: the kernel and Fedora both make the human submitter fully responsible for correctness and licensing.
• Keep it small: focused PRs are easier to trust. Homebrew caps you at one open AI-assisted PR.

If you drive agents from CI, our Claude Code GitHub Actions setup guide covers keeping them inside your own repositories, and Claude Code hooks can block a gh pr create against repos you list.

Warning: Never strip AI trailers or rewrite history to slip AI work past a ban. Asahi and recursion-context treat concealment as grounds for a ban, and it destroys trust with maintainers.

For Maintainers: Write Your Own AI Policy

If AI PRs are flooding your project, a clear written policy saves arguments. Pick one of three common stances and state it in CONTRIBUTING.md.

A short starter you can adapt:

## AI usage
- Do not submit code, docs or issue text generated by AI tools.
- You may use AI for research or debugging if no AI output is in your contribution.
- PRs we believe are AI-generated will be closed without review.
- Questions? Open a discussion before you start.

Then add a checkbox to your PR template, such as “I did not use generative AI to write this contribution”, and link the policy from your README. Submit your repo to osssaysno.com if you want it listed.

Other Trackers and Surveys

• yarikoptic/open-source-ai-contribution-policies on GitHub: a broader list covering bans and permissive policies.
• Slopfree Software Index on Codeberg (brib/slopfree-software-index): projects that reject AI-generated code.
• daily.dev survey (August 2026): of 120 projects studied, 72 allow AI with conditions, 37 ban it and 11 are undecided.

The survey’s numbers matter: bans are a sizeable minority, not the norm. But the list of projects saying no keeps growing, and osssaysno.com’s 22 includes several big names.

What Is Still Unclear

• How projects reliably detect AI-written code; most rely on maintainer judgement.
• Whether courts will settle the copyright status of AI output, which drives the QEMU and Gentoo bans.
• How Asahi will relicense older contributions and whether all past contributors agree.
• Who runs osssaysno.com and how strictly its automated checks verify policies.

My Take

This is less a war on AI than a fight over review time. An agent can open ten PRs in the time a maintainer reviews one, and unpaid maintainers are drawing a line.

For AI users the lesson is simple: treat every project’s AI policy like its licence. Read it, follow it, and disclose honestly where AI is allowed. That is how AI-assisted developers keep their welcome in the projects that still accept them.

For maintainers, the kernel’s Assisted-by: approach looks like the most workable middle ground. It keeps a human accountable without pretending AI tools don’t exist.

FAQ

What is osssaysno.com?

It is a public tracker of open-source projects that refuse AI-generated contributions. Each entry links to the project’s policy and shows when it was last verified.

Which big projects ban AI-generated code?

The tracker lists QEMU, Gentoo Linux, Zig, Servo, Redox OS, immich, Meilisearch, libvips and MonoGame among its 22 projects.

Does the Linux kernel accept AI-generated code?

Yes, with rules. A human must review it, add their own Signed-off-by, take full responsibility, and disclose the tool with an Assisted-by: trailer.

Can I use ChatGPT or Claude just to understand a codebase?

Often yes. QEMU, for example, allows AI for research and debugging if none of the AI’s output ends up in your patch. Stricter projects such as stb reject even AI-reviewed bug reports, so read each policy.

What happens if I submit an AI PR to a project that bans it?

Usually the PR is closed. Some projects go further: recursion-context bans the contributor, and Asahi bans anyone who conceals extensive LLM use.

How do I add my project to the tracker?

Paste your repo’s GitHub URL into the “Suggest a project” form on osssaysno.com. It opens a prefilled issue and runs an automated policy check.

Why did Asahi Linux switch to the GPL?

Its 7.3 progress report says code was taken from personal forks and some was rewritten with AI to hide its origin. It moved to GPL-2.0-or-later to require sharing and attribution.

Sources

• Open Source Says No (tracker)
• QEMU: Code provenance and AI policy
• Linux kernel docs: AI Coding Assistants
• Linuxiac: Fedora opens the door to AI tools
• Gentoo Council AI policy
• Servo AI usage policy
• Asahi Linux LLM policy
• It’s FOSS: Asahi Linux drops MIT for GPL
• daily.dev: AI policies of 120 open-source projects
• GitHub: open-source AI contribution policies list

Related: Claude Code’s rule against cloning existing games.