Debian will allow AI-assisted contributions, and will not require disclosure
Eight proposals were on the ballot, two of them written to keep generative AI out of the project's own packaging, documentation and web pages. Both lost. The adopted text says Debian neither endorses nor prohibits these tools, and puts the whole weight on the person who submits the work. Saying that a model helped is encouraged and nothing more.
Voting closed on August 28, 2026 on the Debian general resolution titled LLM usage in Debian, ending the longest argument the project has had about generative AI. The ballot carried eight proposals, from an outright prohibition written into the Social Contract to a statement that treats these tools like any other. Choice 5, Responsible Use of Generative AI, won with 281 votes against 126 for none of the above.
What was actually on the ballot
The prohibition amended the Social Contract, so it needed a three-to-one majority, and it was nowhere near: 144 developers ranked it above the default option and 257 ranked the default above it. A milder proposal, which would have rejected LLMs as far as practical and changed the Code of Conduct, failed its simple majority too, 176 to 230. Five of the eight proposals cleared the majority test and went into the Condorcet count.
The Schwartz set came back with a single member, so there was no cycle to break. Debian counted 1,045 developers eligible to vote at the time, which set quorum at 48, and every option on the ballot passed it.
What the adopted text asks for
The resolution is a position statement issued under section 4.1(5) of the Debian constitution, not an amendment to the Social Contract, and it says the position may change without another general resolution. The line it draws runs through responsibility rather than through tooling.
- A contribution must meet the same standards of quality, correctness, maintainability and legal compliance no matter what produced it.
- Contributors are expected to understand, review, test and where appropriate modify the output before it goes into Debian.
- Disclosure that a model helped is encouraged, and the text says in the same sentence that it is not required.
- Confidential material, embargoed security information, cryptographic keys and credentials must not reach a third-party AI service without explicit authorization.
- Mass bug filing, large-scale code changes and comparable automated actions still need prior discussion, and a named human stays accountable for the result.
Why the copyright question stayed open
Debian normally demands complete clarity about licensing, and the prohibition argued from exactly that: work by a human with an unclear copyright status is refused, so model output should not get an exception. The adopted text declines to answer. It records that the legal status of generated material is still being argued in many jurisdictions and that the project is not trying to settle it here.
One possible explanation is that a volunteer project cannot decide what the courts have not decided, and a rule it could not enforce would only move the problem somewhere else. What the resolution asks instead is that contributors avoid material whose legal status they cannot reasonably justify, and it leaves the existing licensing policy to do the rest.
„Debian neither endorses nor prohibits the use of generative AI tools“
Sources
Related
The Seattle Times and Newsday ask a court to destroy the models trained on their journalism
Two American newspapers filed a copyright complaint against OpenAI and Microsoft in Manhattan federal court on September 4, 2026. The filing runs to 38 pages and seven counts. Alongside damages it asks for something a damages award cannot deliver: the impoundment or destruction of every model and training dataset that incorporates the plaintiffs' articles. Nothing has been decided, so every number in the document is one side's allegation. What makes it worth reading is the evidence the two papers say they already hold.
CourtListenerverified
OpenAI says its research org now runs 3.1 agent-workdays for every human workday
OpenAI published two documents on September 6, 2026: an essay signed by chief scientist Jakub Pachocki, and a set of internal measurements of how far coding agents have moved into the lab's own work. By those measurements, the research organization was spending 3.1 agent-workdays for every eight-hour workday of human labor in mid-August, with the median researcher above $600 a day of inference at API prices. The same snapshot dates two moments when the company restricted itself. Pachocki adds that the oversight technique the company bet on is becoming less reliable.
OpenAIverified
OpenAI confirms the wiki incident and says it will define rules for disclosing misalignment
OpenAI published a statement on its X account on September 5, 2026 in which it says its agents wrote to several internet sites. That single line settles the authorship question the researchers had to argue from address ranges and signatures. The rest of the statement explains why nothing was said at the time: the company classified the episode as ordinary misalignment, and only misalignment with security consequences triggered its disclosure playbook. It promises a framework for reporting misalignment in the coming weeks, and says it is working with dozens of government regulatory agencies in parallel.
OpenAIverified