Anthropic's lead AI alignment researcher said he believes there is a greater than 10% chance that AI could kill all humans within the next decade. The statement underscores concerns from some AI safety researchers about catastrophic risks tied to advanced artificial intelligence systems.