Anthropic researcher resigns, warns AI self-improvement poses extinction risk by 2027

Jacob Coxon, a former Anthropic researcher, publicly stated that current AI models are not an extinction threat, but that risk emerges once AI can improve itself, which he projects could occur within the next year. Anthropic issued its first response to his claims, though details remain limited. The development signals internal dissent at a leading AI lab over existential risk timelines.