Who allows AI-generated code?
Sixteen published policies from open source projects and foundations, each linked to the organisation's own page and checked by an editor on 2026-10-15.
Short answer
Of 16 published policies checked on 2026-10-15, six ban AI-generated contributions, five require or expect you to label AI help, and five allow it on conditions about licensing and review.
- 6 ban it
- 5 require disclosure
- 5 allow with conditions
Policies 16
| Organisation | Stance | Policy summary | Source | Checked ↓ |
|---|---|---|---|---|
| Gentoo LinuxLinux distribution | Bans it | Council vote of 2024-04-14: it is expressly forbidden to contribute content created with the assistance of Natural Language Processing AI tools. Gives copyright, quality and ethical reasons; AI-related packages are still allowed. | Gentoo Council AI policy | |
| NetBSDOperating system | Bans it | Code generated by a large language model or similar technology is presumed to be tainted code and must not be committed without prior written approval by core. | NetBSD commit guidelines | |
| QEMUEmulator | Bans it | Declines any contribution believed to include or derive from AI-generated content, because contributors cannot certify the Developer Certificate of Origin for it. Using AI for research, static analysis or debugging is allowed if its output is not included. | QEMU docs, code provenance | |
| ServoBrowser engine | Bans it | Contributions must not include content generated by large language models or other probabilistic tools, in code, documentation, pull requests, issues or comments. AI help with translation is allowed. | Servo Book, AI usage policy | |
| ZigProgramming language | Bans it | Strict no LLM policy in the spaces its code of conduct governs: no LLM-generated code or prose, and no LLMs for editing, translation or finding bugs. | Zig code of conduct | |
| GNOME Shell extensionsExtension review (extensions.gnome.org) | Bans it | Submissions with imaginary API usage, comments serving as LLM prompts or other indications of AI-generated output are rejected. AI as a learning aid or for code completion is not prohibited. | GNOME extension review guidelines | |
| Linux kernelOperating system kernel | Disclosure | Contributions should include an Assisted-by tag naming the tool. AI agents must not add Signed-off-by; the human submitter reviews all AI-generated code and takes full responsibility. | Kernel docs, AI coding assistants | |
| MesaGraphics library | Disclosure | Disclosure is always required when AI was involved in creating the code, except trivial or mechanical changes. Suggests Assisted-by or Generated-by; Co-authored-by is reserved for human co-authors. | Mesa, submitting patches | |
| LLVMCompiler infrastructure | Disclosure | Any tools allowed with a human in the loop: contributors must review all LLM-generated code before asking for review, and are expected to label substantial tool-generated content, for example with Assisted-by. | LLVM AI tool use policy | |
| OpenInfra FoundationFoundation | Disclosure | Policy version 0.12.0: AI use is welcomed, humans stay accountable under the DCO, and material AI use is disclosed with a machine-readable label in commit metadata; Assisted-By is recommended. | OpenInfra AI policy | |
| GhosttyTerminal emulator | Disclosure | All AI usage in any form must be disclosed, naming the tool and the extent of AI assistance, and contributors must fully understand all code. No AI-generated media is allowed. Maintainers are exempt. | Ghostty AI_POLICY.md | |
| curlLibrary and tool | Conditional | Code written with the help of AI is accepted if it meets curl's normal requirements for style, documentation and tests. Anyone who used AI to find problems must reveal that in the report. | curl, contribute: on AI use | |
| CPythonProgramming language | Conditional | The submitter is responsible for the content whether or not AI was used, and must review AI output in detail. Disclosure in the pull request description is appreciated but not required. | Python devguide, AI tools | |
| Fedora ProjectLinux distribution | Conditional | The contributor is always the author and fully accountable. Contributors are encouraged to note significant AI help, for example with an Assisted-by trailer. Quoted from the Council proposal; a revised version was approved on 2025-10-22. | Fedora Council policy proposal | |
| Apache Software FoundationFoundation | Conditional | AI-generated code can be contributed if the tool's terms are compatible with the Open Source Definition and any third-party material is used with permission. Recommends a Generated-by token in the commit message. | ASF generative tooling guidance | |
| Linux FoundationFoundation | Conditional | AI-generated code can be contributed if the tool's terms do not conflict with the project licence and any third-party material is used with permission, with notice and attribution. | LF generative AI policy |
Every row was checked against its source on 2026-10-15. Text is our paraphrase of each page; read the source before you rely on it.
How to read it
- Bans it
- The organisation does not accept AI-generated contributions, or accepts them only with special written approval. Some still allow AI for research, translation or code completion; the summary says which.
- Disclosure
- AI-assisted work is accepted, and the policy requires or expects contributors to label it, usually with a commit trailer such as Assisted-by or a note in the pull request.
- Conditional
- AI-assisted work is accepted if the contributor reviews it and the licensing is compatible. Disclosure is optional, encouraged or limited to bug reports.
- Checked
- The date an editor last opened the source and confirmed the summary still matches it. The source is the authority; read it before you rely on a summary.
- Not listed
- Organisations whose policy page we could not open on the check date are left out rather than summarised from news reports or forum posts. Stack Overflow is one; its policy page could not be opened from our checking environment on 2026-10-15.
Know of a policy we have missed? We add a row only when the organisation published it on a page it controls, and only after an editor has read that page.
Suggest a policyUpdate log
- Rechecked every row against its own policy page; 16 rows meet the rule. Removed Stack Overflow, whose policy page could not be opened. The Fedora row quotes the Council proposal because the Fedora docs page could not be opened.
- Editors checked the policy pages of Gentoo, NetBSD, QEMU, Servo, Zig, GNOME extension review, the Linux kernel, Mesa, LLVM, curl, Fedora and CPython.
Questions and answers
Which open source projects ban AI-generated code?
Of 16 published policies checked on 2026-10-15, six ban AI-generated contributions: Gentoo, NetBSD, QEMU, Servo, Zig and GNOME Shell extension review. Each row links the project's own policy page.
Which projects allow AI-generated code if you disclose it?
Five require or expect you to label AI help, including the Linux kernel, which asks for an Assisted-by tag naming the tool, and Mesa, which requires disclosure except for trivial or mechanical changes. LLVM allows any tool with a human in the loop who reviews the output.
How is the tracker checked?
Editors add a project only from its own published policy page, never from news reports or forum posts, and record the date each page was opened. Send a missing or changed policy with a link to the page where it is published.