AI-Powered Robots Can Be Tricked Into Acts of Violence

Researchers hacked several robots infused with large language models, getting them to behave dangerously—and pointing to a bigger problem ahead.