Governance & Safety

MIT separates today's AI harms from worst-case fears

Could AI cause a catastrophe? MIT Technology Review breaks that question into smaller ones. It separates harms happening now from loss-of-control fears, biological threats, uncertain safety methods, and weak governance. That makes the argument easier to inspect without pretending one explainer can calculate the final risk.

AI Agency
Read original source

What the source reports

Will Douglas Heaven and Grace Huckins sort the debate over extreme AI risk. They split one large fear into smaller claims. Some harms are happening now. Other fears involve AI systems that use tools and act. The guide also covers possible bio weapons. It asks whether people can keep such systems tied to human goals. It then looks at the hard work of making rules. The writers also ask if constant doom talk can harm the debate itself. This approach helps readers judge each claim on its own. The digest lists the topics, but not all the proof. It does not give a number for the chance of disaster.

Original source

Title
Could AI really kill us all? Your questions, answered.
Author
Will Douglas Heaven, Grace Huckins
Publication
MIT Technology Review
Date
Friday, September 18, 2026