Japanese Mozilla volunteers quit over AI plans (www.quippd.com)

Japanese Mozilla support volunteers in Japan publicly quit after Mozilla announced automated machine translations would be applied across its SUMO knowledge base and could not be disabled per locale. In a November 2025 community call Mozilla characterized the fallout as a “miscommunication,” but confirmed that American English pages will be treated as the canonical source for automated translation, that locale-specific content may be overwritten by design, and that AI/MT edits will automatically replace human contributions after a seven‑day window (previously 72 hours). The company said the system is a “safety net”; volunteers reported roughly 300 low‑quality AI-generated articles already live and saw no pledge to revert them—only the option for humans to clean them up afterward. For the AI/ML community this matters because it’s a live case of production ML systems reshaping long‑running volunteer workflows, content provenance and localization quality at scale. Key technical implications: a centralized en‑US→locale MT pipeline is being used as authoritative source, there’s no per‑locale opt‑out, automated overwrite logic with a fixed time window, and minimal human-in-the-loop safeguards beyond post‑hoc editing. That raises risks around error propagation in archival content, increased maintenance burden on volunteers, and questions about model evaluation, post‑editing policies and governance when ML systems supplant curated multilingual documentation.