Should organisations prioritise secrets rotation or agent identity design first?

Why This Matters for Security Teams

Security teams often treat secrets rotation as the fastest way to reduce exposure, but that only addresses one failure mode. If the underlying agent identity is shared, overprivileged, or tied to a vague service account, a fresh token simply preserves the same blast radius. The better starting point is to define who or what the agent is, what it may access, and under which conditions. That is especially true when agentic systems can chain tools, request new actions, and behave differently at runtime, which is why guidance from the OWASP Agentic AI Top 10 and NIST AI Risk Management Framework places strong emphasis on governance before control tuning. NHIMG research also shows how quickly identity sprawl becomes operational risk, with the 52 NHI Breaches Analysis illustrating that identity failures are rarely isolated events. In practice, many security teams encounter credential theft only after a design flaw has already made the secret worth stealing.

How It Works in Practice

A practical sequence is to design the identity model first, then automate rotation inside that model. Start by giving each agent a unique non-human identity, not a shared account, and bind it to a specific workload or execution context. Use workload identity where possible, such as cryptographic proof of the workload rather than a static username and password. Then define the minimum scope needed for each task, preferably through intent-based or context-aware authorisation that is evaluated at request time rather than through a fixed role that tries to predict every future action. For autonomous systems, JIT credentials and ephemeral secrets are usually a better fit than long-lived static secrets. Short-lived credentials reduce the time window for misuse, but they only work when the issuing system can revoke them automatically after the task completes. That is why Ultimate Guide to NHIs — Static vs Dynamic Secrets is so relevant here, and why the broader Guide to the Secret Sprawl Challenge focuses on reducing where secrets live, who can reach them, and how quickly they expire. The implementation pattern usually looks like this:

• Assign one identity per agent, per environment, per privilege domain.
• Replace standing access with JIT approval and short-lived credentials.
• Evaluate policy in real time using policy-as-code, not only during provisioning.
• Rotate secrets automatically, but only after scope and ownership are defined.

Current guidance from CSA MAESTRO agentic AI threat modeling framework and the NIST AI Risk Management Framework supports this order of operations. These controls tend to break down when a single agent must impersonate multiple systems across loosely governed pipelines because identity boundaries and revocation paths become ambiguous.

Common Variations and Edge Cases

Tighter identity design often increases delivery overhead, requiring organisations to balance security gains against integration complexity. That tradeoff is real, especially in legacy automation, CI/CD, and batch workloads where teams have historically relied on one shared credential to keep pipelines moving. In those environments, a phased approach is usually more workable than an abrupt overhaul: first split shared identities, then narrow entitlements, then introduce JIT issuance and rotation windows. Best practice is evolving here, and there is no universal standard for every agentic stack yet. The main exception is emergency containment. If a secret is already exposed, immediate rotation may be the right first action to stop active abuse, but that is incident response, not a governance strategy. Another edge case appears in highly dynamic agentic systems where the same model instance can touch multiple tools, prompts, or tenants. In those cases, the control problem is less about secret lifespan and more about ensuring that each action is authorised against current context. The Top 10 NHI Issues and OWASP NHI Top 10 both reflect this pattern: identity misuse becomes hardest to control when one actor is allowed to do too much, too often, for too long. In practice, rotating secrets first helps only when the system already has clear identity boundaries and revocation mechanics in place.

Related resources from NHI Mgmt Group

• Should organisations prioritise secrets rotation or agent approval workflows first?
• Should organisations prioritise secrets rotation or policy controls first for agents?
• When should organisations prioritise Zero Standing Privilege for non-human identities?
• Should organisations prioritise secret rotation or access review first