Non-human identities just outnumbered your employees 144 to 1. J.P. Morgan just confirmed your security perimeter is already compromised.

Your CISO thinks the threat is hackers.

It's not.

J.P. Morgan published data today that should change how every enterprise thinks about security.

Non-human identities — AI agents, service accounts, automated workflows — now outnumber human identities 144-to-1 in enterprise environments.

That ratio grew 44% year-over-year.

And it's accelerating. BeyondTrust's research found a 466.7% increase in AI agents operating inside enterprise environments. Not a typo. Four hundred and sixty-six percent.

Here's what makes this dangerous.

Your identity governance was built for humans. Passwords. SSO. MFA. None of that works for a thousand machine accounts spawning inside your infrastructure every week.

Your SIEM was built to detect human behavior. These agents don't behave like humans.

And 36% of all published AI vulnerabilities now involve APIs — the connective tissue between these agents and your core systems.

72% of U.S. cybersecurity venture deals through May 2026 involved AI-enabled companies. The capital is flowing into the right problems. But the enterprises being protected are still running 2019 identity architecture against a 2026 threat model.

Audit every non-human identity in your environment today. Map their permissions. Kill the unused ones. If you can't answer how many machine accounts you have and what they can access, you don't have a security problem.

You have an identity crisis.

SOURCE: https://www.jpmorgan.com/insights/banking/commercial-banking/ai-cybersecurity-threats-funding-and-builder-priorities
VERIFIED: J.P. Morgan Commercial Banking, BeyondTrust Phantom Labs research, Entro Security H1 2025 report, Wallarm 2026 API ThreatStats report
SIGNAL: The identity perimeter has fundamentally shifted. Enterprises still governing access through human-centric models are exposed to an attack surface 144x larger than they realize.