Prompt injection breaks today’s AI agents, study warns
www.csoonline.com
Researchers say current AI agents fail to consistently resist prompt injection attacks, exposing enterprises to failures that conventional security testing may overlook.