The lead maintainer of GNU Emacs was asked on the project’s public development list whether he would knowingly accept code produced by slave labour. He replied:
“Our answer must be YES.”
The question came from brian cully on 6 November 2026, in the thread on Human Emacs, a group of contributors who seek an editor free of LLM-generated code. Cully’s hypothetical satisfied every formal requirement.
- A territory where slavery is lawful.
- Title to the product of forced labour granted by statute.
- Copyright assigned to the FSF.
- Code technically correct.
- Maintainers fully informed of how it was made.
Read those carefully. Then look at the answer. Would acceptance be ethical?
Our answer must be YES. Any other answer is a first step towards splitting our community along the many lines of political and ideological divides that have nothing to do with software freedoms.
That is ethically reprehensible, and also logically false.
Eli Zaretskii has maintained Emacs since 2016. He has been asked this same question repeatedly lately, three times over three days.
On 8 November cully wrote that this amounted to tolerating slavery for the benefit of Emacs. Cully is right.
Zaretskii replied: “I don’t tolerate slavery. I just don’t ask our contributors how they feel about slavery.” He added that “people rarely agree what is injustice nowadays. Someone’s terrorist is another one’s freedom fighter. Someone’s justified war is another’s genocide.”
Eli is wrong. The hypothetical had stipulated that the maintainers already knew.
Later that day Divya Ranjan Pattanaik asked about child slavery. Zaretskii answered with a question of his own:
Are you okay with accepting code from countries whose policies you condemn or whose regimes you consider cruel and inhuman? If yes, why is that different from the slavery example above?
That exchange shows what the argument is about for Eli. Cully asked about the provenance of an artifact: how a piece of code was made. The question was about the code. Zaretskii answered each time about persons, and in the third reply he named which persons: contributors from countries whose regimes others condemn. An ethics test would reach countries, so Eli wants to stay blind to everything except licence paperwork and code “quality”. He stated the rule himself:
we will not change our principles, regardless of the way the code we accept was written.
Hume described this move in 1740, so this whole debate lands nearly three centuries late. Moralists reasoned in “is” for pages and then shifted into “ought” as if the two were the same kind of statement. Zaretskii makes that shift in a single line. The assignment is signed, the patch compiles, therefore the project must accept it. Legality is an “is.” Obligation is an “ought.” The second has to be argued for on its own, and across three messages the argument offered is that asking would be divisive.
Wollstonecraft answered the divisiveness claim even more clearly in 1790. Burke had argued that inherited institutions must stay beyond ethical examination, because a society that judges its own foundations fractures. Her reply applies to “fracture our community into many different enclaves” word for word: “the whole tenor of his plausible arguments settles slavery on an everlasting foundation.” In her second edition she spelled out the consequence. On Burke’s reasoning, “the slave trade ought never to be abolished.”
Sounds familiar, no?
A rule that forbids ethical examination of provenance protects whatever provenance is already inside the system. That held for sugar. It holds for emacs.
The GNU Project has written guidelines for tone. Richard Stallman posted to the same thread on 4 November to remind a participant of the Kind Communication Guidelines. His last post predates the slavery answer by two days. Philip Kaludercic, who maintains ELPA and argued for a middle path on LLM code, replied at length to Pattanaik and passed over the child slavery paragraph in silence. The project currently rejects LLM-generated code when a contributor discloses it, pending advice from FSF lawyers. Provenance is already factored in. The dispute is over which facts about provenance count. Legal title counts. Coerced labour is ruled out of scope.
Wollstonecraft called Burke’s position a servile reverence for antiquity. Shole would call Emacs yes to slavery a reverence for copyright assignment.