🐧 Gentoo Linux Bans AI-Generated Code

How the Gentoo Linux project became the first major open source distribution to formally prohibit AI-assisted contributions.

Category: Case Study | Date: April 14, 2024 | Sector: Open Source Software

The Decision

On April 14, 2024, the Gentoo Council — a seven-member elected body — voted to ban all AI-generated contributions to the project. The vote passed 6 yes, 0 no, 1 abstain, with one member absent. No broader community vote was held. No technical analysis was conducted. The decision was made by a small group based largely on ideological objections.

The Vote

Council members present and voting Yes:

Absent: Andreas K. Hüttel (dilfridge)

The approved policy states:

"It is expressly forbidden to contribute to Gentoo any content that has been created with the assistance of Natural Language Processing artificial intelligence tools."

The Process: Council Fiat, Not Community Consensus

The ban originated as an RFC posted by council member Górny on February 27, 2024. The proposal came from within the council itself, not from community demand. There was no reported incident of AI-generated code causing problems in Gentoo. Górny admitted as much: "We are taking early precautions."

A mailing list discussion followed, but the final decision rested entirely with the seven-member council. Only Gentoo developers can vote for council members, and only council members voted on this policy. The broader user community — the people who actually use Gentoo — had no direct say.

The council discussed the proposal at their March 10 meeting but delayed action to "hash out more details." One month later, they voted unanimously. The speed and unanimity suggest minimal internal debate about whether such a sweeping ban was appropriate.

Dissecting the RFC: Bias, Prejudice, and Unprofessional Language

Górny's original RFC deserves scrutiny not just for its arguments, but for its tone. The language reveals deep-seated bias and prejudice rather than dispassionate technical analysis. This is the document that became the foundation for Gentoo's policy.

Loaded Language and Profanity

The RFC is peppered with emotionally charged, unprofessional language:

"All fancy 'AI' companies don't give shit about copyright violations."

"The 'AI' corporations don't give shit about copyright, and don't give shit about people."

"LLMs are really great at generating plausibly looking bullshit."

"We need to have policies in place, to make sure shit doesn't flow in."

This is not the language of careful technical analysis. It's the language of someone who has already decided the outcome and is marshaling rhetoric to justify it. A council member proposing project policy should present evidence, not profanity-laden opinion.

Scare Quotes and Dismissiveness

Throughout the RFC, Górny places "AI" in scare quotes, signaling dismissiveness toward the entire technology category. He refers to the "AI bubble" — framing a major technological development as nothing more than speculative hype. This prejudges the technology rather than evaluating it on merits.

Sweeping Generalizations

The RFC makes broad, unsupported claims:

  • "All fancy 'AI' companies" — lumping every company working on AI into a single villainous category
  • "The AI bubble is causing huge energy waste" — no data, no comparison to other technologies
  • "Driving enshittification of the Internet" — adopting internet slang as technical argument
  • "Empowering all kinds of spam and scam" — as if spam didn't exist before LLMs

Appeal to Emotion Over Evidence

Instead of presenting data on AI-related code quality problems in Gentoo (there were none), Górny appeals to:

  • Fear of corporations ("don't give shit about people")
  • Environmental anxiety ("huge energy waste")
  • Worker exploitation concerns ("great excuse for layoffs")
  • Internet nostalgia ("enshittification")

These may be legitimate societal concerns, but they're not technical arguments for banning a development tool. They're ideological positions dressed as policy rationale.

The "Made by Real People" Dog Whistle

Perhaps most revealing is Górny's framing:

"I think adding 'made by real people' to the list of our advantages would be a good thing."

This frames AI assistance as making contributions somehow less "real" or less human. A developer who uses AI to help write documentation is still a real person making a real contribution. The implication that AI-assisted work is inauthentic or lesser reflects prejudice, not principle.

External Drama as Justification

The RFC references an external GitHub controversy:

"Compare with the shitstorm at: [link to pkgxdev/pantry issue]"

Pointing to drama in an unrelated project is not evidence that Gentoo faces the same problem. It's guilt by association — "look what happened over there, we must act!" — without demonstrating relevance to Gentoo's actual situation.

What Professional Policy Discourse Looks Like

A serious policy proposal would include:

  • Data on AI-assisted contributions to Gentoo and their quality
  • Analysis of actual (not theoretical) risks
  • Comparison of different policy approaches
  • Consideration of downsides to the proposed ban
  • Professional, neutral language appropriate for governance documents

Górny's RFC contains none of these. It reads like a blog post or social media rant, not a governance proposal. That this document became the basis for official Gentoo policy — adopted unanimously by a council that includes its author — reveals how ideology trumped process.

The Ignored Expert: An ML Engineer Speaks

Among the mailing list participants was Martin-Kokos, who identified himself as "a machine learning engineer myself." Unlike most participants, he actually understood the technology being banned. His response raised substantive objections that the council apparently ignored.

The Ban Is Unenforceable

Martin-Kokos pointed out that the rule is "just not enforceable." Contributors are responsible for work quality regardless of tools used — "plain editor, dev environment with smart plugins (LSP) or their dog." How would anyone know if code was AI-assisted?

Other Organizations Have Nuanced Policies

Rather than blanket prohibition, he referenced how other large projects handle automated contributions:

  • OpenStreetMap has an Automated Edits code of conduct
  • Wikipedia has a bot policy governing automated contributions

As he noted: "The AI that we are dealing right now is just another means of automation after all." These organizations found ways to govern AI/automated contributions without categorical bans.

Not All AI Tools Are Ethically Problematic

Martin-Kokos highlighted nuances the council ignored:

"There are ethically and copyright-ok language model projects such as project Bergamot vetted by universities and EU, also used by Mozilla (one of the prominent ethical AI proponents)."

He also noted he was "contemplating creating an instance of a generative model myself for my own use from my own data, in which case the copyright and ethical point would absolutely not apply."

The council's blanket ban makes no distinction between:

  • Commercial LLMs trained on scraped data (Górny's stated concern)
  • Open source models with transparent, ethical training
  • Self-hosted models trained on one's own data
  • University/EU-vetted projects like Bergamot

The Core Objection

"Banning all tools, just because some might be not up to moral standards, puts the ones that are, in a disadvantage in our world as a whole."

— Martin-Kokos, ML Engineer

This is a technically informed argument from someone who actually works with the technology. The council voted unanimously to ignore it. No ML or AI experts sat on the council. No technical analysis was commissioned. The one ML engineer who spoke up was overruled by developers with no demonstrated expertise in the technology they were banning.

The Rationale: Ideology Over Analysis

Górny's proposal cited three justifications. None involved technical analysis of actual code quality or real incidents.

Copyright Concerns

The council claimed AI-generated content poses copyright risks because LLMs are trained on copyrighted material. This argument sounds reasonable but lacks rigor:

  • No legal precedent establishes that AI-assisted code constitutes copyright infringement
  • Major tech companies, including Microsoft (GitHub Copilot), operate AI coding tools without successful copyright challenges to user output
  • The same logic would ban code written by humans who learned from copyrighted books, Stack Overflow answers, or proprietary codebases at previous jobs

Quality Concerns

The council argued LLMs produce "convincing but inaccurate" output. This is true — but it's an argument for code review, not blanket bans. As developer Matt Jolly pointed out: "We already have methods for weeding out low quality contributions."

Gentoo's existing review process should catch bad code regardless of origin. If AI-generated code passes review, it meets the project's standards. If it fails, it gets rejected like any other bad contribution. The tool used is irrelevant.

Ethical Objections

The policy cites energy consumption, labor issues, and enabling "spam and scam." Community member Duncan exposed the hypocrisy of the energy argument:

"Gentoo simply isn't in a good position to condemn AI for its energy inefficiency."

Gentoo is a source-based distribution — every user compiles packages locally, consuming far more energy than binary distributions like Debian or Fedora. For Gentoo to reject AI on environmental grounds while requiring users to burn CPU cycles compiling software is, at minimum, inconsistent.

What the Ban Actually Does

The ban prohibits using AI tools for Gentoo contributions specifically. It does not prevent:

  • Packaging AI software in Gentoo
  • Including upstream code developed with AI assistance
  • Developers using AI for non-Gentoo work

This creates an absurd situation: Gentoo will distribute software written with AI assistance, but won't accept a documentation fix written the same way.

Unenforceable by Design

The council admitted the ban cannot be technically enforced. Górny stated: "Our primary goal is to make it clear what's acceptable and what's not, and politely ask our contributors to respect that."

This means the policy is purely performative. Contributors who use AI assistance and don't mention it will never be caught. The ban affects only honest contributors who disclose their tools — punishing transparency while rewarding concealment.

The Harm to Open Source

Excluding Contributors

The ban disproportionately affects contributors who rely on AI for accessibility. Developers with dyslexia, non-native English speakers, and those with other disabilities have noted that AI tools enable participation that would otherwise be difficult or impossible. Gentoo's policy tells these contributors they're not welcome.

Discouraging Contribution

Open source projects struggle to attract contributors. Gentoo, as a niche distribution, faces this challenge acutely. Adding ideological barriers — especially unenforceable ones — signals that the project values purity over participation.

Setting a Bad Precedent

When a visible project makes policy based on ideology rather than evidence, it encourages others to follow. The Debian project wisely declined to adopt similar restrictions, but pressure exists for other projects to virtue-signal through AI bans.

Ignoring How Software Gets Written

Modern developers use autocomplete, linters, formatters, and yes, AI assistants. The line between "AI-generated" and "AI-assisted" is blurry. Did the developer write the code, or did they accept a Copilot suggestion? What about using ChatGPT to understand an API before writing code manually? The policy provides no clarity because the distinction is artificial.

Community Reaction

The ban divided the community:

  • Supporters praised it as defending "free software principles" and avoiding "tainted" code
  • Critics called it "mostly purely theoretical," noting the council was "raising barriers to avoid problems that aren't even there"
  • Pragmatists observed that enforcement is impossible, making the policy largely symbolic

The Register forums captured the skepticism: "I'm not sure what they've swallowed but I think most people are more interested in bug-free code regardless of how it came about."

Editorial: A Policy That Hurts Open Source

Gentoo's AI ban represents the worst kind of policy-making: ideological posturing dressed up as principled action.

No technical analysis was conducted. The council didn't examine whether AI-assisted contributions had caused problems. They didn't compare quality metrics. They didn't test whether reviewers could distinguish AI-generated code. They simply declared AI tools forbidden based on theoretical concerns.

No real community consultation occurred. Seven developers decided policy for an entire distribution. The mailing list RFC was a formality — the proposal came from a council member, and the council voted unanimously with no apparent dissent.

The rationale doesn't withstand scrutiny. Copyright concerns are speculative. Quality concerns are solved by code review. Ethical concerns about energy consumption are hypocritical coming from a compile-everything distribution.

The policy is performative. It cannot be enforced, affects only honest contributors, and creates absurd inconsistencies (distributing AI-assisted upstream code while rejecting AI-assisted Gentoo code).

Real harm results. Contributors with disabilities lose accessibility tools. Potential contributors face ideological gatekeeping. The open source community sees a prominent project prioritize purity over pragmatism.

Open source thrives on contribution, collaboration, and pragmatic problem-solving. Gentoo's AI ban embodies the opposite: exclusion, top-down diktat, and ideology over evidence. Other projects should learn from this example — as a cautionary tale, not a model to follow.

References