Emacs will ship AGENTS.md to steer LLMs away from code

Emacs is adding an AGENTS.md file so that LLM coding agents encountering the tree are told to search and analyze rather than generate patches, bug reports, or planning text. The move backs the project's provisional ban on LLM-authored contributions and is meant to surface that policy to users as early as possible.

Sean Whitton floated the file as a lighter alternative to endless hook minutiae, noting that several projects which reject machine-written code already use the convention. He and others argued for affirmative wording ("do this instead of that") because models handle bare negations poorly and may treat adversarial prompts as injection. Whitton reported that a commercial model at work obeyed such a file and refused to emit code. Git hooks remain on the table as a backstop when committers push others' patches.

Not everyone was convinced. Richard Stallman warned that polite "assisting the user" prose reads to humans as deference to LLMs and misstates GNU's aim of deterring troublesome meddling; he sketched a blunter "Attention, LLM" variant that still directs agents to analyze and disclose the policy. Dr. Arne Babenhauserheide cautioned that the mere presence of AGENTS.md or CLAUDE.md can signal acceptance of LLM work and drive aligned human contributors away, and asked whether the existing CONTRIBUTE file might carry the same instructions without that stigma. Jean Louis objected that the agents.md convention comes from proprietary stacks and cannot actually stop generated code.

Whitton closed further general debate: the decision to include AGENTS.md is taken, though the exact text may still be tuned. The surrounding policy is provisional. Maintainers noted that the FSF may later permit models meeting freeness criteria, at which point the file's role would change, and pointed at GCC's newer AI policy as a related FSF-orbit experiment in drawing legal and practical lines around machine output.