The 2026 Infrastructure Identity Survey done by Teleport, an AI Infrastructure Identity Company, reveals that while 92% of enterprises have moved AI into full-scale production, security is dangerously lagging. This “preparedness gap” has already led to AI-related incidents for 60% of companies. The report identifies a critical risk in “over-privileging,” noting that 70% of organizations grant AI systems more access than human employees, making them 4.5 times more likely to suffer security breaches.
A “confidence paradox” has also emerged: leaders who feel most secure in their AI deployments actually report incident rates double those of their less confident peers. This stems from a reliance on legacy models like static credentials, which 67% of firms still use despite a 20% increase in associated risk. As AI shifts from a tool to an autonomous “actor,” governance remains nearly non-existent, with only 3% of organizations using automated real-time prevention.
With over half of infrastructure leaders expecting AI to run major operations autonomously within three years, the report argues for a fundamental shift. Decision-making power is already moving toward platform and infrastructure teams, who must now transition from vulnerable passwords to identity-native, cryptographic security to contain the risks of agentic AI.
