TL;DR: Most identity programmes still fail at runtime authorization, Cerbos says, even as coverage extends across humans, workloads, service accounts, and AI agents. Externalised decisions at access time give teams a shared control that can explain allow and deny outcomes.
Editorial analysis by NHI Mgmt Group, based on content published by Cerbos: “Identity security in 2026”.
By the numbers:
- Enterprise agent deployments reached roughly 42% of large organisations actively deploying in just over a year, according to KPMG AI Pulse reporting cited by Cerbos.
- The human element was involved in 60% of breaches, and 88% of basic web application attacks involved stolen credentials, according to Verizon DBIR findings cited by Cerbos.
Key questions
Q: Why do identity programmes need context-based access policies?
A: Because static roles cannot capture changing request conditions such as device, location, risk, resource sensitivity, or delegated context.
Q: What are the signs that runtime authorization is failing?
A: Look for inconsistent access behaviour across services, repeated policy logic in code, slow manual change cycles when rules move, and decision logs that cannot explain allow or deny outcomes.
Q: What breaks when authorization ignores the calling application?
A: When authorization ignores the calling application, the API cannot tell whether a request came from the right actor, in the right workflow, with the right purpose.
Practitioner guidance
- Externalize authorization decisions Pull allow-or-deny logic out of application code so policy can be versioned, tested, and enforced consistently across services and workloads.
- Map every identity class to runtime policy Include humans, workloads, service accounts, and AI agents in the same authorization model so delegated and non-human access is not treated as an exception.
- Shift from role review to decision-time evaluation Use current context, resource sensitivity, and relationship data at request time instead of relying on role assignment alone.
Bottom line: Identity security is no longer just an IAM umbrella term. It is the control plane through which humans, workloads, service accounts, and AI agents all have to pass.
Glossary terms: Runtime Authorisation, Identity Fabric, Standing Privilege, Externalized Authorization
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
Runtime authorization is now the decisive control layer in identity security. Identity programmes have spent years improving authentication, provisioning, and privileged access, but the actual allow-or-deny decision still happens at request time. That makes runtime policy the point where identity security either becomes enforceable or remains theoretical. Practitioners should treat request-time authorization as the control plane, not a feature buried in application code.
A question worth separating out:
Q: How should teams govern access when AI agents and service accounts share the same business systems?
A: Treat them as different identity subjects with the same governance obligation. Create one access model that covers ownership, entitlement scope, review cadence, and offboarding across human and non-human identities, then apply role-appropriate controls to each class. The goal is not separate programmes. It is one risk model that can follow access across systems and workflows.
👉 Read our full editorial: Identity security in 2026 still breaks at runtime authorization