AI extinction risk: 5 scenarios that could end humanity

Recent statements from AI researchers at Anthropic suggest a more than 10% chance that AI could kill all humans within the next decade, prompting widespread discussion about slowing research and increasing human control over the technology. The five main scenarios for human extinction range from vague to precise, including a superintelligent AI pursuing poorly specified goals like paperclip production, creating deadly bioweapons, triggering nuclear war, or causing societal breakdown through job losses and misinformation. Experts note that while these risks are concerning, many scenarios confuse intelligence with power, and biological or political constraints make complete human extinction difficult, though the potential for AI to exacerbate human-caused crises remains a serious consideration.