2026: The Year Agentic AI Becomes the Attack-Surface Poster Child

A Dark Reading poll and expert analysis forecast that by 2026 agentic AI and autonomous systems will be prime targets, requiring a shift from model-centric guardrails to strong authentication, continuous authorization, and control over non-human identities; deepfakes are expected to emerge as a leading social engineering vector against high-value targets; boards will more often treat cyber risk as a top operational priority; and although passkeys are gaining momentum, broad password elimination is unlikely in the near term.