RFC: Banning "AI"-backed (LLM/GPT/whatever) contributions to Gentoo

Enforceability and Detection

• Many doubt the ban can be enforced: there’s no reliable way to tell if text or code was AI‑assisted, especially when humans edit it.
• Some argue the point isn’t detection but policy: if something looks AI‑generated and low‑quality, maintainers have explicit grounds to reject it without further effort.
• Others warn this risks arbitrary decisions or “witch‑hunts,” since style can be mimicked or hidden.

Impact on Quality and Maintainers

• Strong concern that LLMs enable huge volumes of superficially plausible but incorrect code, docs, and bug reports.
• Example discussed: autogenerated package descriptions that confidently misrepresent software, plus a bot giving useless replies.
• Maintainers fear the time ratio flipping: contributors can generate patches in minutes that require deep human review, overwhelming volunteers.
• Some say you should always judge by quality; AI or human, the review burden is the same. Opponents reply that zero‑cost generation makes spam qualitatively different.

Role and Value of LLMs

• Many report concrete benefits: autocomplete‑like coding (Copilot), faster navigation of large codebases, better API search than Google/Stack Overflow, test generation, and code review ideas.
• Repeated warning: LLMs hallucinate, invent APIs, misread specs, and must be checked via docs and tests.
• Debate on whether AI should be banned entirely, or only in code review / automated submissions.

Scope and Definitions of “AI‑backed”

• Significant confusion over what counts: LLMs vs translation tools (e.g., DeepL) vs other ML‑based features.
• Some propose a distinction: AI that generates new content vs AI that transforms user‑authored content, but others note both share similar failure modes and underlying tech.
• Unclear lines risk inconsistent enforcement.

Legal, Ethical, Energy, and Security Concerns

• Some worry about copyright contamination from training on protected code and about “code laundering.”
• Others argue learning from public code is aligned with open‑source values, provided verbatim copying is avoided.
• Energy and water usage of LLMs is criticized; counter‑arguments note Gentoo’s own compile‑from‑source model is also resource‑heavy.
• Security and trust: with incidents like the xz backdoor as backdrop, some advocate treating AI‑produced contributions as higher‑risk and not worth reviewer time.

Reputation and Community Dynamics

• Concern that visible AI use (even just icons or descriptions) damages project reputation; users may infer the entire project is AI‑generated and low‑effort.
• Some see the ban as conservative but strategically bold; others view it as impractical, overbroad, and driven by personal dislike of AI rather than targeted quality controls.