TL;DR: Identity visibility collapsed from 93% in 2024 to 46% in 2025 in Permiso Security’s 2026 State of Identity Security Report, while 92% of organisations already have AI agents accessing production or sensitive data and 95% say those systems can create or modify identities without traditional oversight. The programme failure is no longer inventory quality but real-time visibility into who or what is acting.
At a glance
What this is: This report shows identity visibility falling sharply while AI agents, service accounts, and other non-human identities expand access across production environments.
Why it matters: IAM, NHI, and security teams need to treat visibility as a live control surface, because static inventories no longer explain who or what can act inside the environment.
By the numbers:
- Comprehensive identity visibility fell from 93% in 2024 to 46% in 2025.
- 92% already have AI agents accessing production or sensitive data.
- 95% said AI systems could create or modify identities without traditional human oversight.
Context
Identity visibility is the ability to see what identities are doing in real time, not just to list them. Permiso Security’s 2026 State of Identity Security Report argues that many organisations still confuse inventory with visibility, which leaves them unable to track access, permissions, and blast radius fast enough to matter.
The report’s core point is that the problem has shifted from counting identities to governing their runtime behaviour. That matters for NHI programmes because service accounts, API keys, tokens, and AI agents can now create an operational gap between provisioning and oversight that traditional IAM reporting does not close.
The findings are drawn from responses from more than 500 security and identity professionals worldwide. The pattern described in the article looks less like an isolated maturity issue and more like a broad control-plane failure across cloud identity operations.
Key questions
Q: What breaks when identity visibility stops at inventory?
A: Inventory without attribution leaves teams unable to tell which accounts, entitlements or non-human identities still have a valid business purpose. That gap lets stale access persist long after ownership changes, role changes or workload changes. Effective governance requires continuous validation, not just discovery, because a visible identity is not necessarily a justified one.
Q: Why do AI agents make non-human identity governance harder?
A: AI agents make governance harder because they can request tools, act autonomously, and change behaviour across sessions while still relying on machine credentials. That increases the number of access paths security teams must supervise. The result is a stronger need for task-scoped access, explicit ownership, and continuous monitoring of what the agent can reach.
Q: How do teams know if identity security controls are actually working?
A: Identity security controls are working when teams can show a current view of high-risk entitlements, detect privilege drift quickly, and remove access before exposure spreads. A useful sign is reduced time between entitlement change and policy review. Another is fewer unresolved conflicts between approved access and actual production permissions.
Q: What should organisations do when service accounts keep accumulating after systems change?
A: They should connect discovery to offboarding so stale service accounts, API keys, and tokens are not left active after the application or owner changes. The issue is not just accumulation, but continued successful authentication. Lifecycle closure needs to be a live control, not a periodic clean-up exercise.
Technical breakdown
Why inventory is not the same as identity visibility
An inventory tells you what identities exist. Identity visibility tells you what those identities are doing, where they are acting, and how much damage they could cause if compromised. The article’s distinction matters because lists are static while access behaviour is dynamic, especially in cloud environments where permissions, sessions, and resource paths change quickly. Teams that stop at discovery can pass a compliance check and still miss active exposure. Practical implication: build visibility around runtime identity activity, not just periodic asset lists.
Practical implication: build visibility around runtime identity activity, not just periodic asset lists.
Why AI agents turn identity visibility into a control problem
The article describes AI agents that access sensitive data and can create or modify identities without traditional oversight. That changes the governance model because the system is not just authenticating a subject, it is allowing an identity-like actor to trigger new access states during execution. In practice, this compresses the time available to detect drift, review privilege, or explain who authorised what. For IAM teams, the issue is not only access at provisioning, but access created inside the workflow itself. Practical implication: treat AI-driven identity creation as a governed runtime event, not a back-office admin action.
Practical implication: treat AI-driven identity creation as a governed runtime event, not a back-office admin action.
Why unified cross-platform visibility is the architectural gap
Permiso Security’s report shows organisations using multiple tools to piece together identity activity, with no single source showing the complete picture. That fragmentation matters because IdPs, CSPMs, and SIEMs each expose a slice of identity state, but none alone can show permissions, usage, and blast radius together. The operational result is slow correlation and delayed response during identity compromise. A mature identity programme needs a control plane view that joins discovery, entitlement, and activity. Practical implication: align identity telemetry so permissions and behaviour can be analysed together in one operational view.
Practical implication: align identity telemetry so permissions and behaviour can be analysed together in one operational view.
Threat narrative
Attacker objective: The attacker aims to exploit valid identity paths to reach sensitive data or critical systems while remaining inside trusted access flows.
- Entry occurs when AI agents or other non-human identities are granted production or sensitive-data access without adequate runtime oversight. The article shows that these subjects already operate inside many environments, which turns access into a live governance issue rather than a future risk.
- Escalation happens when those identities can create or modify additional identities without traditional human approval. That behaviour expands the access surface faster than manual review cycles can track, especially where service accounts and API keys are provisioned at scale.
- Impact follows when teams cannot determine blast radius quickly enough after suspicious identity activity is detected. The article ties delayed visibility to longer response time and weaker containment across cloud environments.
Breaches seen in the wild
- Scania insurance portal breach 2025: An attacker used an external user login, likely stolen by infostealer malware, to take insurance claim documents from a Scania portal.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Visibility is now the primary control plane for identity security. The report shows a sharp drop in confidence because organisations are finally measuring what they can actually observe, not what they assume exists. That shift matters across human, NHI, and agentic programmes because a static inventory never told the whole story about runtime access. The practitioner conclusion is simple: if you cannot see identity behaviour in motion, you do not have control.
The visibility illusion cascade is a programme design failure, not a tooling problem. Security teams often assemble IdP, CSPM, and SIEM data and still lack a joined view of identity permissions and activity. That is not a dashboard issue, it is a governance failure to define where identity truth lives across systems. The practitioner conclusion is to treat visibility as a control architecture, not a reporting layer.
AI agents expose an assumption that identity creation is human-paced and reviewable. That assumption was designed for actors whose access changes through tickets, approvals, and auditable handoffs. It fails when the actor can create or modify identities during execution without traditional oversight. The implication is that access governance must be rethought around runtime issuance, not post-hoc certification.
Unified identity telemetry is becoming the dividing line between mature and fragile programmes. The article’s numbers show that organisations can detect identity problems within a day but still struggle to determine blast radius in minutes. That gap is where response slows and compromise spreads. The practitioner conclusion is that identity telemetry must support containment decisions, not just investigation after the fact.
Credential graveyards are an outcome of confidence without lifecycle control. When organisations believe they know their NHI estate but still rely on periodic discovery or manual correlation, expired and unused credentials keep authenticating in production. That is a lifecycle governance problem with security consequences, not a housekeeping issue. The practitioner conclusion is that offboarding and discovery have to be operationally linked.
From our research library:
- 96% of security operations teams report critical blind spots, most commonly in cloud infrastructure (74%) and identity and access behaviour (67%).
- Read next: Identity Security Programme Guide
What this signals
Identity visibility now behaves like a control plane issue rather than a reporting issue. Teams that still treat visibility as a quarterly inventory exercise will continue to miss runtime identity behaviour, especially where AI agents and service accounts can change access states faster than review cycles can react. The practical shift is from asking what identities exist to asking what they are doing right now.
Real-time identity visibility is becoming the minimum viable condition for effective response. The article shows why disconnected telemetry creates delay at the exact point where blast radius must be understood. 96% of security operations teams report critical blind spots, most commonly in cloud infrastructure and identity behaviour, according to the Ultimate Guide to NHIs. That reinforces the need to join identity, cloud, and logging data into one decision surface.
For practitioners
- Separate inventory from visibility in your IAM programme Define inventory as a discovery output and visibility as a runtime control outcome. Track whether you can answer who is acting, what they can reach, and how quickly blast radius can be determined when identity behaviour changes.
- Map AI agent identity creation to approval boundaries Identify every place where AI systems can create or modify identities, tokens, or permissions without a human approval workflow. Treat those paths as governance exceptions until they are explicitly owned and monitored.
- Unify identity telemetry across core platforms Correlate IdP authentication data, cloud identity data, and SIEM access events into one operational view so permission scope and activity can be analysed together instead of in separate tools.
- Review NHI discovery and offboarding together Tie service account discovery to lifecycle closure so unused or expired identities are revoked, not merely reported. Prioritise credentials that still authenticate successfully after their business owner has moved on.
- Measure blast-radius time as a control metric Track how long it takes to determine which systems, data, and identities are exposed after an alert. Use that measure to test whether your visibility layer is actually helping containment.
Key takeaways
- The report shows a growing gap between perceived identity control and actual runtime visibility across cloud environments.
- AI agents and other non-human identities are expanding access faster than many governance processes can observe or certify it.
- Organisations need unified telemetry, faster blast-radius analysis, and lifecycle controls that work at execution time, not only after review.
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 and MITRE ATT&CK address the attack and risk surface, while NIST CSF 2.0 and CIS Controls v8 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-02 — Secret Leakage | The report highlights exposed and unmanaged identities that remain active without visibility. |
| NHI-05 — Overprivileged NHI | The article ties weak visibility to unclear permissions and blast radius across identities. | |
| NHI-07 — Long-Lived Secrets | Expired and unused credentials stay active in production, which is the article's lifecycle risk. | |
| Recommendation — Scan for leaked or untracked NHI credentials and revoke any that still authenticate successfully. Reduce NHI privilege scope and verify access boundaries against actual runtime usage. Shorten secret lifetimes and enforce revocation when identities are no longer needed. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The article focuses on valid identity abuse and the blast-radius that follows. |
| Recommendation — Map identity compromise to credential access and lateral movement hunting priorities. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The report is fundamentally about whether permissions and entitlements are actually visible and governed. |
| DE.CM-01 — Networks and Networks Monitoring | Identity visibility depends on continuous monitoring of relevant activity signals across environments. | |
| Recommendation — Continuously validate entitlements against observed identity behaviour and remove excess access. Instrument identity telemetry so monitoring can detect changes in access behaviour quickly. | ||
| CIS Controls v8 | CIS-5 — Account Management | The credential graveyard and unmanaged identities point directly to account lifecycle control. |
| Recommendation — Centralise account inventory, disable stale identities, and verify owner accountability. | ||
Key terms
- Identity Visibility: Identity visibility is the ability to see which identities exist, what they can access, and how those access paths relate across systems. In NHI programmes, it means correlating service accounts, tokens, certificates, and agents into one operational view so governance decisions are based on evidence, not assumptions.
- Blast Radius: The potential scope of damage if a specific credential or identity is compromised. Identities with broad permissions have a larger blast radius and represent a higher priority for least-privilege enforcement and security controls.
- Credential Graveyard: A credential graveyard is the accumulation of expired, unused, or orphaned secrets that still authenticate in production. These stale service accounts, API keys, and certificates create hidden access paths that survive long after their original purpose has ended.
- Runtime identity governance: Runtime identity governance is the discipline of checking identity behaviour while access is being used, not just when it is granted or reviewed. It combines telemetry, policy comparison, and response so organisations can detect when access drifts from intent across distributed systems.
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 June 24, 2026.
Updated on October 11, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org