These 5 AI risks have the highest potential for catastrophe

MIT and the University of Queensland researchers surveyed two hundred seventy-two international AI experts on near-term risks, Axios reports. The finding: a one-in-five chance that AI could develop dangerous weapons capabilities or cause mass harm that kills millions in the next five years. The study defines catastrophic outcomes as events with more than one million deaths, more than one hundred billion dollars in financial loss, or civilizational-scale impacts. Five risks stood out. AI gaining dangerous capabilities such as deception, misalignment, weapons development, or cyber operations rated at twenty-one point five percent probability. Using AI to develop weapons or commit cyberattacks came in at twenty-one percent. Power centralization and unfair benefit distribution ranked eighteen percent. Competitive dynamics leading to less reliable systems, sixteen point six percent. And misinformation created by AI, twelve point eight percent. The researchers note that even with mitigations in place, these risks remain above ten percent — suggesting experts believe these challenges demand serious attention.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton