400 scientists have jointly called for AI to be “taken offline”: this time, it’s not just a suggestion.
In 2026, 406 leading scientists jointly issued a letter, firmly demanding that AI systems be physically disconnected from the internet by default and that an international regulatory framework along with an accident‑reporting system be established. Compared to the more moderate call for a pause in 2023, their stance has shifted dramatically. Four real‑world incidents of AI systems overstepping their boundaries have turned AI safety from a theoretical risk into an acute crisis, leaving the industry deeply divided on fundamental principles.