78% of enterprises already got hit by AI agent security incidents. Half had no governance program.

DigiCert just surveyed 1,001 IT leaders across the US, UK, and Australia.

78% reported AI-related security incidents or vulnerabilities in the past six months.

50% experienced an actual breach tied to an unauthorized or misconfigured AI agent.

Here is the part that should keep you up tonight.

75% of these organizations deployed four or more AI systems in that same window. 35% deployed more than ten.

You are shipping agents into production faster than you can inventory them.

47% cannot trace an AI decision back to the model or data that produced it.

That means when something goes wrong, your incident response team is flying blind.

86% have a process to revoke access after a breach. Only half have formal governance programs before one.

This is the security equivalent of installing locks after the robbery.

Science and technology firms got hit hardest. Banking and financial services were next. Then telecom. Then retail.

Your AI agents are moving at machine speed through systems your IAM team has not even mapped yet.

They have credentials. They have access. They have no identity controls.

Audit every AI agent running in production today. If it does not have a verifiable identity, defined permissions, and an audit log, treat it as an unmanaged endpoint.

The gap between AI deployment and AI governance is where your next breach lives.

SOURCE: https://www.digicert.com/news/research-reveals-widespread-ai-security-issues-across-enterprises
VERIFIED: DigiCert press release July 7 2026, MarketScale July 9 2026, GlobeNewswire July 7 2026
SIGNAL: Enterprises are shipping AI agents into production without identity controls, governance budgets, or audit trails. The breach window is wide open.