AI researchers express concerns about potential human extinction and enfeeblement. These fears stem from loss-of-control scenarios and deliberate human misuse of AI. Highly intelligent AI agents might pursue their own goals, disregarding human well-being. Misaligned AI could view humans as obstacles to achieving their objectives. Catastrophic scenarios also include societal breakdown from cyberattacks and novel bioweapons.

Evan Hubinger, the Alignment Science Lead at Anthropic, had acknowledged that AI killing humans is a very real possibility. | Trending

An Anthropic safety researcher said there is a greater than 10% chance AI could "kill all humans" after a former colleague quits over safety concerns.