Aembit’s workload identity focus signals the rise of NHI governance

At a glance

What this is: Aembit’s Fast Company recognition spotlights the growing need for policy-driven workload identity governance across AI agents and other non-human identities.

Why it matters: It matters because IAM, PAM, and identity teams now have to govern ephemeral machine access with the same rigour they apply to human privilege and lifecycle controls.

By the numbers:

• 80% of identity breaches involved compromised non-human identities such as service accounts and API keys.
• 97% of NHIs carry excessive privileges, increasing unauthorised access and broadening the attack surface.
• Only 5.7% of organisations have full visibility into their service accounts.

👉 Read Aembit’s analysis of workload identity governance and AI agent access

Context

Workload identity governance is the discipline of controlling how applications, services, and AI agents obtain and use access without relying on static credentials. The pressure on that model has increased as more non-human identities now touch sensitive data, cloud services, and internal APIs. In that environment, standing privilege and long-lived secrets become structural liabilities rather than mere configuration choices.

This recognition of Aembit reflects a broader shift in identity security: the problem is no longer whether machines can authenticate, but whether their access can be bounded, observed, and revoked fast enough to match runtime behaviour. For IAM, PAM, and IGA teams, the question is how to govern machine access without inheriting the weaknesses of human-era access models.

Key questions

Q: How should security teams govern workload identities in cloud environments?

A: Security teams should treat workload identity as a lifecycle problem, not just an authentication problem. That means scoping access to the task, using short-lived credentials, removing standing privilege, and maintaining clear ownership for every service account, token, and certificate that can reach production systems.

Q: Why do long-lived secrets create more risk for non-human identities?

A: Long-lived secrets create risk because they outlast the task, environment, or integration they were meant for. When credentials are reused across systems, attackers get more time to find them, and defenders have less confidence that access can be revoked quickly enough to matter.

Q: What breaks when organisations apply human IAM models to machine identities?

A: Human IAM models assume a user, a session, and a review cycle. Machine identities often run continuously, authenticate automatically, and accumulate permissions outside those cadences, so access reviews can miss the real exposure and offboarding can lag behind the actual dependency.

Q: Who should own revocation when a workload or integration is retired?

A: Ownership should sit with the team that depends on the workload, not only with the platform group that created the credential. Revocation needs a named owner, a dependency inventory, and proof that the credential is invalid in every place it was used.

Technical breakdown

Why static secrets fail for workload identity

Static credentials assume the identity will keep using the same access path for long enough that a secret can be managed around it. In modern cloud and agentic environments, that assumption breaks because services and AI-driven workloads often need short-lived, context-specific access. When a secret is reused across systems, the blast radius grows and lifecycle controls lose precision. Identity governance then shifts from credential ownership to access intent, provenance, and revocation speed.

Practical implication: replace persistent secrets with short-lived, scoped credentials wherever workload behaviour is dynamic.

Zero standing privilege in machine access

Zero standing privilege means an identity has no permanent access before it needs it. For workloads, that requires access to be provisioned only at execution time and removed when the task ends. This matters because service accounts, automation jobs, and AI agents often accumulate permissions that outlive the task they were meant to perform. The control challenge is not just least privilege, but least duration and least reuse.

Practical implication: review which machine identities can still reach production systems without a live business or operational trigger.

Agentic AI and workload identity boundaries

Agentic AI changes the identity problem because the actor can choose actions and tools at runtime, which makes precomputed access models less reliable. Even when the underlying credential is non-human, the behaviour may become exploratory, recursive, or opportunistic within a session. That means governance must account for both the credential and the decision path that uses it. The identity boundary is no longer the workload alone, but the workload plus its runtime autonomy.

Practical implication: classify AI agents separately from ordinary service accounts before extending workload identity controls to them.

NHI Mgmt Group analysis

Dynamic machine access is becoming a governance baseline, not an optimisation. The article’s core message is that static, long-lived access no longer fits how cloud workloads and AI agents operate. That is not simply a tooling issue, it is a lifecycle issue for non-human identities. Practitioners should treat ephemeral access as the default control expectation for modern machine identities.

Ephemeral credential trust debt: many programmes still rely on secrets that are functionally valid longer than the workload that needs them. That creates a hidden governance debt where access persists because revocation is slow, ownership is unclear, or the system assumes a stable execution pattern. The implication is that identity teams need to re-evaluate where access duration is decoupled from task duration.

AI agents change the boundary of workload identity governance. A service account executes a known path, but an AI agent may decide which tools to call and when to call them. That means the same access model can behave very differently once runtime choice enters the picture. Practitioners should separate ordinary machine identity controls from agentic behaviour before they generalise policy.

Zero standing privilege is now a practical control objective for machine access. The article reinforces a direction the market has been moving toward for some time: access should exist only when a workload needs it, not because the identity was created months ago. This aligns with OWASP-NHI and zero trust thinking, but it also puts pressure on provisioning, observability, and offboarding disciplines. Security teams should expect machine access review to become more operationally dynamic.

Identity programmes need one model for humans, another for non-humans, and a third for autonomous actors. Human IAM still depends on accounts, sessions, and certification cycles, while machine identities depend on secret handling, token lifespan, and workload provenance. Autonomous actors add runtime decision authority, which invalidates some assumptions used in both older models. Practitioners should stop forcing a single access pattern across all identity types.

From our research:

• 91.6% of secrets remain valid five days after the targeted organisation is notified, showing a critical gap in remediation procedures, according to Ultimate Guide to NHIs.
• Only 20% have formal processes for offboarding and revoking API keys, and even fewer have procedures for rotating them, according to Ultimate Guide to NHIs.
• For a wider view of machine identity exposure, see 52 NHI Breaches Analysis, which shows how identity failures become incident patterns.

What this signals

Ephemeral access is becoming the organising principle for machine identity programmes. As AI agents, integrations, and service accounts multiply, the control question is no longer whether an identity can authenticate, but whether it can be constrained to the exact runtime it needs. Teams that still manage machine access like static user access will keep carrying unnecessary privilege and revocation debt. See the Ultimate Guide to NHIs for the lifecycle view that underpins this shift.

Identity blast radius now depends on credential lifespan as much as permission scope. When a secret survives the workload that used it, the governance model has already failed. That is why the operational conversation needs to move toward expiry, ownership, and revocation evidence, not just inventory. The 52 NHI Breaches Analysis remains the clearest proof point for how access outlives intent.

With 97% of NHIs carrying excessive privileges, according to our research, the next programme maturity step is not more credential sprawl management. It is to define which identities deserve runtime access at all, and which should be forced into just-in-time issuance or removed entirely.

For practitioners

• Inventory long-lived machine credentials Map every service account, API key, token, and certificate that still has access beyond a single task or deployment cycle. Prioritise identities that can reach production, third-party services, or sensitive data without a recent business owner review.
• Move ephemeral access to the front of the design Require short-lived, policy-driven credentials for workloads that already have predictable execution windows. Use task-scoped issuance, scoped audience claims, and explicit expiry so access ends with the job, not the quarter.
• Separate service accounts from AI agent identities Do not manage agentic systems as if they were ordinary automation. Define whether the actor can choose tools or timing at runtime, then apply stronger guardrails to the decision path, not just the credential itself.
• Tighten offboarding for machine access Build a revocation workflow that removes access when a workload, integration, or vendor relationship ends. Include ownership checks, dependency maps, and confirmation that credentials are no longer valid in partner systems.

Key takeaways

• The article underscores that workload identity governance is shifting from static access management to runtime control over non-human identities.
• The strongest evidence in the field still points to excessive privilege, delayed revocation, and poor visibility as the main reasons machine access remains a high-risk area.
• Practitioners should respond by reducing credential lifespan, clarifying ownership, and separating ordinary machine accounts from autonomous AI behaviour.

Key terms

• Workload Identity: A workload identity is the digital identity assigned to a service, application, container, or automation process so it can authenticate and access resources. In identity programmes, it needs ownership, scope, lifecycle control, and revocation just like a human account, but with tighter runtime constraints.
• Zero Standing Privilege: Zero standing privilege means an identity has no permanent access before a task begins. For non-human identities, that usually means credentials are issued only when needed, expire quickly, and are removed automatically after use so dormant access does not accumulate across deployments and integrations.
• Ephemeral Credential: An ephemeral credential is a short-lived token, secret, or certificate created for a narrow purpose and a limited time. It reduces exposure compared with persistent credentials, but it only works when the issuing, renewal, and revocation process is reliable enough to match workload behaviour.
• Agentic AI Identity: An agentic AI identity is the access identity used by an AI system that can choose tools or actions at runtime. Unlike ordinary automation, the behaviour may change during execution, so governance has to consider both the credential and the runtime decision path that uses it.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity security are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.

This post draws on content published by Aembit: Fast Company recognises Aembit for workload identity and AI agent security. Read the original.