TL;DR: Akeyless reports that only 7% of organisations believe current controls would stop a compromised agent from operating, because credential protection alone does not govern what happens after authentication across machines, workloads, and autonomous systems. Runtime authority and zero standing privilege matter because AI-era access risk is now action-level, not just secret-level.
At a glance
What this is: Akeyless argues that AI agent identity security now needs runtime authority because static access controls and credential protection do not stop misuse after authentication succeeds.
Why it matters: IAM, PAM, and NHI programmes need to shift from protecting credentials to governing actions, because autonomous systems can still misuse valid access even when secrets are short-lived.
By the numbers:
- Only 7% of organisations believe their current controls would actually prevent a compromised agent from operating.
- Akeyless' 2026 State of AI Agent Identity Security report is based on a survey of 400 IT and security leaders.
- Research cited by Akeyless says machine identities now outnumber human identities by more than 140 to 1.
- Akeyless says it secures over 220 billion machine identity interactions.
👉 Read Akeyless's analysis of AI agent identity security and runtime authority
Context
AI agent identity security is not just a secrets problem. Once a machine, workload, or autonomous system authenticates, the real risk is whether anything can constrain what it does next inside production systems, data stores, and SaaS platforms.
Akeyless frames the gap as a shift from credential management to operational control. That matters because ephemeral credentials reduce exposure windows, but they do not by themselves stop a valid identity from taking an action that exceeds intent.
The article is grounded in a 2026 survey of 400 IT and security leaders and in incident patterns involving trusted machine access, OAuth permissions, service accounts, and agent credentials. The underlying governance problem is now broader than any one identity type.
Key questions
Q: What breaks when AI agents keep standing credentials?
A: The access model breaks because the agent can continue acting after the human has moved on, the workflow has shifted, or the original approval is no longer relevant. Standing credentials turn delegated authority into unattended authority, which is especially risky when agents can retry, chain tools, and move quickly across systems.
Q: When should organisations prioritise runtime authority over more secret rotation?
A: Prioritise runtime authority when authenticated identities can still cause damage after login, especially for AI agents, workloads, and privileged service accounts. Rotation reduces exposure to stolen secrets, but it does not stop misuse of valid access. If the question is post-authentication behaviour, runtime control comes first.
Q: What are the signs that machine identity management is failing in an organisation?
A: Common signs include incomplete inventory, spreadsheet based tracking, manual renewal processes, unclear ownership, and repeated certificate expiry events. Another signal is when teams struggle to audit where machine identities exist or cannot automate lifecycle actions at scale. If operational teams keep reacting to expiring certificates instead of governing identity lifecycles proactively, the control environment is already under strain.
Q: How should teams govern AI agent access to production data?
A: Treat agent actions as runtime events that need policy checks at execution time, not just pre-approved credentials. High-risk reads, writes, and deletes should be bounded by context, session limits, and approval gates that apply before the command completes.
Technical breakdown
Why static credentials no longer define the control point
Traditional IAM assumes that access is mostly decided at issuance and then monitored later. That model fits persistent users and long-lived service accounts, but it breaks when cloud workloads, APIs, and AI agents operate continuously and change context during execution. The security issue is no longer merely whether a secret can be stolen. It is whether the identity can still do harm after legitimate authentication has already occurred. Short-lived credentials reduce exposure, but they do not automatically constrain the action space once access exists.
Practical implication: shift security design from secret custody to runtime control over what authenticated identities are allowed to do.
How zero standing privilege changes machine identity governance
Zero Standing Privilege removes permanent access and replaces it with task-scoped, time-bound permissions. In this model, a credential should exist only when a workload or agent needs it, and it should disappear immediately after use. That is a stronger fit for modern machine identity estates because dormant permissions are what attackers abuse most often. ZSP also changes the operational question from whether a secret has rotated to whether the access window ever existed outside a specific task.
Practical implication: eliminate standing machine permissions where task-scoped issuance is feasible, especially for production and third-party access.
What runtime authority adds that authentication does not
Runtime authority is an enforcement model that evaluates action intent while a session is active, not just identity at the front door. In the article's example, an AI agent may authenticate successfully and still attempt a destructive database action. Runtime authority inserts a policy checkpoint between request and execution, so the control decision is based on the action being attempted, the task context, and the approved scope. That is the key architectural difference: authentication proves identity, runtime authority governs behaviour.
Practical implication: place policy enforcement where sensitive actions execute, not only where identities log in or fetch credentials.
Threat narrative
Attacker objective: The attacker wants to turn trusted access into operational control over data, infrastructure, or customer systems while avoiding traditional authentication barriers.
- Entry occurs through legitimate authentication by a machine, workload, or AI agent using valid but potentially standing access.
- Escalation happens when the identity can act beyond its intended task because the session still carries broad permissions or dormant authority.
- Impact follows when attackers, prompt injection, or misuse leverage that valid access to reach data, modify systems, or trigger destructive actions without needing to break authentication.
Breaches seen in the wild
- Dropbox Sign breach 2024: A compromised back-end service account gave attackers Dropbox Sign customer data, including API keys, OAuth tokens and MFA information.
- OneLogin API flaw (CVE-2025-59363): A OneLogin API flaw exposed OIDC client secrets to anyone with an API key, including vendors (CVE-2025-59363); fixed with no customer impact.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 150+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Runtime authority is the missing control plane for AI agent identity security: static credentials answer who can enter, but they do not answer what the identity is allowed to do once inside. That distinction is now central because AI agents and machine workloads can chain actions at runtime faster than human review cycles can react. Practitioners should treat action control as a first-class identity function, not a monitoring afterthought.
Zero standing privilege is becoming the baseline assumption for machine identities: persistent permissions create dormant blast radius, and dormant blast radius is what modern attackers exploit. When identities are granted only for a specific task and expire immediately after use, the attack surface shrinks from a credential estate to a governed execution window. The implication for identity teams is that standing access should now be the exception, not the design default.
Ephemeral credential trust debt: short-lived secrets reduce exposure, but they do not remove the governance debt created when organisations assume valid access equals safe behaviour. That assumption was designed for slower, human-paced access reviews and fails when an AI agent can acquire, use, and discard authority within one task. The implication is that review-based governance must yield to issuance-time and execution-time controls for autonomous systems.
Machine identity scale is no longer a side issue in IAM planning: research cited by Akeyless says machine identities now outnumber human identities by more than 140 to 1, which means the control problem has already shifted to the non-human estate. IAM, PAM, and NHI teams should stop treating agent access as a niche extension of user governance. It is now the dominant operational surface in many environments.
Platformisation is a response to lifecycle fragmentation, not a feature preference: the article's argument reflects a real organisational problem, where vaults, PAM, certificates, secrets, and AI controls are split across separate tools. Fragmentation makes it difficult to enforce one policy across human identities, service accounts, and AI agents. Practitioners should interpret the market movement toward unified identity control planes as a governance necessity, not a packaging trend.
From our research library:
- 80% of identity breaches involved compromised non-human identities such as service accounts and API keys, according to the Ultimate Guide to NHIs.
- 73% of vaults are misconfigured, leading to unauthorised access and exposure of sensitive data, according to the Ultimate Guide to NHIs.
- Read next: Guide to NHI Rotation Challenges
What this signals
Runtime authority changes the control question for autonomous systems: identity teams can no longer rely on the old sequence of authenticate, authorise, then review. When a machine or agent can complete meaningful work inside one execution window, the governance model has to constrain the action itself. That is why issuance-time controls and execution-time policy enforcement will matter more than access reviews alone.
Runtime authority changes the control question for autonomous systems: identity teams can no longer rely on the old sequence of authenticate, authorise, then review. When a machine or agent can complete meaningful work inside one execution window, the governance model has to constrain the action itself. That is why issuance-time controls and execution-time policy enforcement will matter more than access reviews alone.
For practitioners
- Replace standing machine permissions Inventory service accounts, API keys, and agent credentials that remain valid outside a specific task, then phase out any access that is not issued just in time for execution.
- Move policy checks to execution time Require sensitive actions to pass through a runtime enforcement point so intent, task context, and approved scope are checked before the action reaches production systems.
- Separate credential custody from agent execution Ensure autonomous agents do not hold reusable secrets, tokens, or passwords directly, and broker access only through ephemeral issuance channels.
- Review third-party and OAuth access paths Map external applications, SaaS integrations, and delegated permissions that can reach internal resources, then remove any broad authorisations that outlive the business task.
Key takeaways
- Static credentials are not the only problem. The larger issue is that valid machine access can still be abused after authentication succeeds.
- Akeyless cites a 7% confidence level among 400 IT and security leaders, which shows how weak current controls look against compromised AI agents.
- Runtime authority and zero standing privilege matter because they move governance from credential possession to action-level control.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10, OWASP Agentic AI Top 10 and MITRE ATT&CK address the attack and risk surface, while NIST CSF 2.0 sets the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-04 — Insecure Authentication | The article focuses on authentication that does not govern post-login agent behaviour. |
| NHI-05 — Overprivileged NHI | Broad, persistent permissions are the central risk the article calls out. | |
| NHI-07 — Long-Lived Secrets | The article contrasts standing credentials with ephemeral identity and runtime control. | |
| Recommendation — Apply NHI-04 controls to ensure authentication is paired with runtime enforcement for machine identities. Reduce overprivileged NHI access by issuing task-scoped permissions that expire after use. Replace long-lived secrets with short-lived credentials and brokered issuance wherever possible. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | The article's core concern is an agent using valid authority beyond intended scope. |
| Recommendation — Treat agent privilege abuse as a runtime control problem and enforce action-level policy before execution. | ||
| MITRE ATT&CK | TA0006;TA0040 — Credential Access; Impact | The article describes trusted access being abused to reach impact without classic exploitation. |
| Recommendation — Map machine identity abuse to credential access and impact techniques, then harden the paths attackers inherit. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article is about governing what an identity may do after authentication. |
| Recommendation — Enforce PR.AA-05 by scoping entitlements to task context and revoking standing permissions. | ||
Key terms
- Runtime authority: Runtime authority is the permission an AI system has while it is actively deciding and acting, not just when it is approved. In governance terms, it is the point where access, tool use, and action scope become operational, which is why build-time review alone cannot prove safety.
- Zero Standing Privilege: A control model in which an identity does not keep persistent access unless it is actively needed. For NHIs, this means credentials and permissions are issued for a narrow task and then removed. It reduces the time window and reuse value of stolen access.
- Dynamic Ephemeral Identity: Dynamic Ephemeral Identity is a model in which credentials or authority exist only for a short operational window and are generated at runtime. It reduces the value of exposed secrets, but only if the environment can also limit what the identity is allowed to do while active.
- Machine Identity: The digital identity of a machine, device, or workload, such as a server, container, or VM, used to authenticate it within a network. Sometimes used interchangeably with NHI, though NHI is the broader category.
What's in the full article
Akeyless's full article covers the operational detail this post intentionally leaves for the source:
- How runtime authority brokers requests before credentials are issued to an agent
- Architectural detail on zero credentials on the agent and zero direct connectivity
- Examples of containerised runtime enforcement close to production systems
- Deployment patterns for combining ephemeral identity, ZSP, and policy checkpoints
Deepen your knowledge
NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.
Published by the NHIMG editorial team on July 9, 2026.
Updated on October 7, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org