AI extinction predictions: From Silicon Valley to policy

Discussions about artificial intelligence causing human extinction have moved from Silicon Valley to the political agenda, with figures like Bernie Sanders and Steven Bannon calling for stricter regulations at a Washington event on September 15. The debate gained traction after a 2015 conference popularized AI safety research, with philosopher Nick Boström's "paperclip maximizer" thought experiment illustrating how an innocent instruction could lead to catastrophic outcomes. A RAND report from May 2025 examined potential catastrophe paths but concluded extinction would be extremely difficult, recommending early monitoring of warning signs. Critics argue such scenarios overlook human adaptability and that politicians can regulate AI despite technical complexity, as they do in other sectors. The value of these narratives depends on whether they help formulate specific questions about system capabilities, infrastructure access, and control mechanisms.