Palo Alto Networks’ 2026 State of Cloud Security Report, based on a survey of 2,800 security professionals, finds that security controls lag AI-driven development. It says 73% of organizations let AI agents modify code, pipeline configurations or production environments, while 19% of high- and critical-severity issues reach production per release cycle.
Palo Alto Networks’ report says AI agents support autonomous or semiautonomous development actions at 99% of organizations, while 73% let agents modify code, pipeline configurations or production environments. Organizations enforce fewer than three of eight assessed predeployment security checks, and 19% of high- and critical-severity security issues reach production per release cycle.
Moving a code fix from detection to production takes 18.4 days on average, the report says, and only 19% of organizations have unified visibility across development and cloud environments. Meanwhile, 65% of security leaders consider frontier AI-driven vulnerability discovery and exploitation a severe threat to critical infrastructure within 12–18 months.