Survey of 209 Infosec Pros Flags 55% AI Phishing Rate, 49% Data-Leak Fears
Updated
Updated · O'Reilly Media · Oct 21
Survey of 209 Infosec Pros Flags 55% AI Phishing Rate, 49% Data-Leak Fears
3 articles · Updated · O'Reilly Media · Oct 21
Summary
A survey of 209 security professionals found AI-generated phishing was the most commonly observed attack at 55%, ahead of vulnerability scanning and exploitation at 30% and AI-assisted malware generation at 26%.
Risk perceptions diverged from what respondents saw in the wild: 49% cited overreliance on AI without human verification and 49% cited sensitive-data leakage as top concerns, while only 25% named AI phishing and 11% prompt injection.
Agent-related worries centered on governance rather than model exploits, with 28% pointing to excessive permissions or unclear ownership, 19% to weak monitoring, and 18% to inability to verify agents stayed within role.
Policy quality tracked closely with safer use and funding: 72% said their company has an AI policy, but only 44% called it comprehensive; where policy was comprehensive, 17% reported widespread shadow AI and 49% said AI security was sufficiently funded.
Adoption remains skewed toward general-purpose tools—68% use chat or copilot tools for research or documentation, versus 38% for vulnerability scanning—and 34% said their biggest skills gap is using AI tools more effectively.