VentureBeat’s June 2026 Pulse Research, surveying 107 enterprises, found that 69% run AI agents with shared credentials somewhere in their deployments. A single compromised agent inherits the permissions of every workflow the shared key touches, leaving no forensic trail to identify the source of an attack.
This vulnerability has triggered a buying spree in enterprise security. Palo Alto Networks completed its $21.1 billion acquisition of CyberArk on February 11, the largest deal in the firm’s history. CrowdStrike closed a $740 million acquisition of SGNL and by June 15 shipped Continuous Identity for AI Agents, the first product from that deal.
The findings underscore a fundamental security gap: most enterprises lack the identity infrastructure to manage individual AI agent permissions at scale. Without per-agent credentialing, attackers can move laterally across workflows undetected.
Cisco has also invested heavily in this space, though exact figures were not detailed in the research. The trend signals a shift toward identity-based security architectures designed specifically for autonomous agent fleets, rather than traditional perimeter defenses.
The research suggests that enterprises should prioritize agent-specific identity management. However, implementing such systems at scale remains complex, and the survey’s sample size of 107 enterprises may not represent the full market.