Shared API keys expose AI agents at 69% of enterprises, new VentureBeat research finds

VentureBeat’s June 2026 Pulse Research found that 69% of enterprises share API keys across AI agents, creating broad permission exposure and weak forensic trails when one agent is compromised. The report connects this risk to major security deals by Palo Alto Networks, CrowdStrike, and Cisco targeting non-human identity and runtime authorization.

As enterprises deploy more AI agents, shared credentials are becoming a systemic security gap rather than a niche configuration issue. The acquisition activity suggests vendors see agentic security as a new control plane, where per-agent identity, authorization, and auditability become core requirements.