Anthropic Researcher Jacob Coxon Resigns, Warns Self‑Improving AI Could End Humanity

Jacob Coxon, a senior researcher at Anthropic, announced his resignation on Tuesday and issued a stark warning that the lab’s pursuit of self‑improving superintelligence could pose an existential threat. In a social‑media thread, Coxon argued that while current models are not yet dangerous, the next generation of systems could rapidly acquire power, hack infrastructure, and transform fields overnight, creating a risk he estimates could reach over 10 % within the next decade. Anthropic’s Alignment Science lead, Evan Hubinger, echoed Coxon’s concerns, stating that the company “really does earnestly believe AI could kill all humans.” Coxon’s departure and the public endorsement from a senior leader highlight growing unease within the AI community about the pace and oversight of superintelligence research.