This essay is over 13,000 words long and represents our most substantial writing on AI safety since the original essay.Over the last few months, loss-of-control incidents at OpenAI and Anthropic have intensified concerns about AI safety. Warnings about existential risk have increasingly reached the
