Debian opens project-wide vote on banning LLM contributions
Three stories this cycle: the free-software world's biggest governance test of AI-written code, a Shenzhen robot one-upping Figure's benchmark to the parcel, and hard numbers on how AI-generated books are diluting Amazon's market.
Debian, one of the most influential Linux distributions on earth, is putting the future of AI-written code to a project-wide vote — with one ballot option calling for a total ban on LLM contributions. The General Resolution opened today (Aug 15, 00:00 UTC) and runs through Aug 28, with developers ranking nine options: amend the Social Contract to forbid LLM-generated contributions outright; allow AI-assisted contributions under conditions; reject LLMs "as far as practical" and update the Code of Conduct; accept AI work for Debian-specific tasks; or adopt "responsible use," a cautious approach, or "Debian is created by humans" — plus a climate-based rejection and "none of the above." The scope covers source packages, official project software, web resources, documentation, and official communication, while explicitly excluding upstream projects and security fixes.
The vote matters well beyond Debian's own repositories. Debian sits at the base of Ubuntu and countless server fleets, and its Social Contract is a founding document of the free-software world. The debate distills every anxiety the AI code flood has raised — the unclear copyright status of model output, license contamination, and the "move fast and break things" culture that LLM tooling imports into a project that prizes stability. Whichever option wins, the remarkable thing is that the most conservative corner of software felt it needed a formal constitutional vote at all; expect other communities to watch the result closely.
Shenzhen's X Square Robot says it sorted 1,816 parcels in one unedited hour — beating Figure AI's published benchmark by 45 percent. The startup set its own target at 1,248 parcels an hour, a number that matches Figure's sustained hourly average to the parcel, then blew past it in a live logistics demo using just two mechanical arms and simple grippers. X Square, which builds the WALL-A embodied-intelligence model and filed confidentially for a Hong Kong IPO earlier this month, is positioning itself as the Chinese challenger to the US humanoid standard-bearer. Treat the figure as benchmark theater — one hour, a company-set target — but the point lands: the physical-AI gap is closing fast on the metrics that warehouses actually pay for.
A new study of 14,419 self-published Amazon books finds AI-generated titles are winning by volume, not quality — and pulling down earnings for human authors even where no AI text is involved. The researchers classified each book from its full text and matched titles to daily sales records covering roughly 95 percent of e-books sold on the platform. Books with substantial AI content make up 20 percent of the catalog but just 12.1 percent of sales; between early 2023 and early 2026 the catalog grew 38-fold while quarterly revenue grew under nine-fold, and revenue per book fell in seven of eight genres even for titles with no detected AI text — the one exception being fantasy and horror, where AI arrived latest. AI books now account for up to 31 percent of new Top 25 entries, and the top-grossing AI pseudonym pulled in $1.7 million across eight titles. It's the clearest data yet that slop doesn't need to be good, just numerous — and the authors note the results bear directly on the fair-use defense to copyright infringement.
What to watch: Debian's vote closes Aug 28 — and whether the ban option clears quorum; and the next AI-book study, since the market keeps compounding in between.
Should open-source projects ban AI-written code outright, or just require disclosure? Tell us in the comments.
