TL;DR: As AI tools spread across departments, legacy IT becomes a control and visibility bottleneck, with JumpCloud citing that 37% of IT professionals see unauthorized access by automated agents as a serious threat and more than 50% of enterprises say legacy IT slows scaling. Old identity and device foundations no longer match the pace of agentic adoption.
At a glance
What this is: This is an analysis of how legacy IT creates governance and visibility bottlenecks for agentic AI adoption, with a focus on identity, device and permission control.
Why it matters: It matters because IAM, IGA and security teams need unified control over human users and automated tools before departmental AI adoption outpaces governance.
By the numbers:
- Over 50% of enterprises find legacy IT actively slows their ability to scale.
Context
Legacy infrastructure becomes a governance problem when identity, device and application control are fragmented across old workflows and siloed data. In an agentic AI environment, that fragmentation does not just slow operations. It weakens the organisation's ability to see which tools are acting, who approved them and what permissions they are using.
JumpCloud's central argument is that AI adoption is now exposing the limits of legacy IT faster than many security programmes can adapt. The article frames unified identity and device management, open interoperability and Zero Trust as the practical response to departmental AI adoption, but the deeper issue is that old control planes were built for slower change and clearer ownership.
Key questions
Q: How should security teams govern AI use in developer tooling?
A: Security teams should govern AI use as a data and access problem, not only a productivity feature. Define what information can be sent to models, require human review of generated code, and apply least privilege to connected repositories and tools. Approved use cases should be explicit, monitored, and revisited as model capabilities expand.
Q: Why do legacy systems create more risk as agentic AI spreads?
A: Legacy systems create risk because they fragment ownership, permissions and audit trails across tools that were never designed to coordinate at machine speed. When AI tools can act with little human intervention, those fragments become blind spots that hide privilege misuse and make scale harder to govern.
Q: What are the signs that AI governance controls are not keeping pace with adoption?
A: Common warning signs include unclear ownership for AI use cases, inconsistent approval processes, limited visibility into where sensitive data enters models, and weak evidence for audits or assessments. Teams also struggle when privacy, security, and legal review happen late or manually, because that usually means governance is reactive rather than embedded in the AI delivery process.
Q: How should organisations balance interoperability with permission control in agentic AI?
A: Treat interoperability as a governance requirement, not just an integration feature. Standard protocols make it easier for AI tools to collaborate, but that collaboration must still be constrained by authorisation, logging and scope review so the organisation can see and limit what each tool can do.
Technical breakdown
Why legacy identity sprawl breaks agentic AI governance
Legacy environments usually accumulate duplicate identities, manual approvals and inconsistent entitlements across tools that do not share a common control plane. That structure can still function for human workflows, but it becomes brittle when AI tools spread department by department and interact with sensitive systems at machine speed. The problem is not only access volume. It is that identity, device and application state are no longer synchronised well enough to support reliable oversight. Practical implication: governance teams need a single view of identity and device trust before agentic usage expands.
Practical implication: consolidate identity and device visibility before AI tooling spreads further across departments.
What Model Context Protocol and A2A change for security teams
Model Context Protocol, or MCP, and Agent2Agent, or A2A, are standardised communication layers that help AI tools exchange context and cooperate across vendors. From an identity perspective, that matters because standardisation increases interoperability, but it also increases the need for consistent authorisation, logging and scope control across tool chains. Without those controls, the same openness that improves utility can widen the blast radius of a mis-scoped agent. Practical implication: treat interoperability as an access design problem, not only an integration problem.
Practical implication: tie interoperability projects to authorisation, logging and scope review rather than treating them as integration only.
Why autonomous tooling exposes permission design defects
The article's risk signal is that AI tools can now act with little human intervention, which makes permission design more consequential than in conventional SaaS adoption. When departments introduce their own tools, the organisation can lose track of which identities are human, which are automated and which are acting with delegated authority. That is where privilege escalation risk grows. The control gap is not the existence of AI itself, but the absence of a permission model that keeps pace with how quickly these tools are deployed and reconfigured. Practical implication: define and review permissions at the pace of tool adoption, not annual cycle time.
Practical implication: review delegated permissions whenever AI tools are added, expanded or repurposed.
Threat narrative
Attacker objective: The objective is to obtain and abuse access through poorly governed AI tooling, then move data or actions through identities the organisation cannot fully see.
- Entry occurs when departments adopt AI tools locally, outside central visibility, creating unmanaged access paths into sensitive systems.
- Escalation follows when automated agents interact with data and services using permissions that were never reviewed for machine-like usage.
- Impact appears as privilege misuse, hidden data movement and slower scaling because the organisation cannot reliably see or control the tools now operating inside the environment.
Breaches seen in the wild
- JumpCloud breach 2023: North Korean hackers breached JumpCloud and abused its device commands framework against a few customers; all admin API keys were reset.
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
Legacy IT has become a governance bottleneck, not just an efficiency drag: The article is right to frame old infrastructure as a control problem because fragmented data, manual workflows and siloed tools make it hard to know what is connected, who approved it and what it can reach. That is exactly where agentic adoption collides with identity governance. The practitioner conclusion is simple: if visibility is weak at the foundation, AI scale will amplify the weakness rather than mask it.
Unified identity and device control is now part of AI governance: This is no longer just endpoint hygiene or SSO administration. When AI tools are deployed through departments, the organisation needs one permission model that spans humans, services and automated tools. Without that unification, entitlement drift becomes invisible until a tool is already operating beyond its intended scope. The implication is that IAM and device governance must be designed together, not handed off to separate programmes.
Legacy infrastructure creates identity blast radius debt: That named concept captures the hidden cost of old systems in an agentic environment. Each fragmented permission set, local approval path and disconnected audit trail expands the damage an automated tool can do before anyone notices. The debt is not theoretical; it is the accumulation of all the places where access decisions cannot be traced back to a single governance model. Practitioners should treat blast radius as a measurable property of infrastructure maturity.
Open interoperability only works when authorisation keeps up: The article's emphasis on MCP and A2A is directionally correct, but interoperability without identity discipline just moves risk faster. Standardised communication makes it easier for AI systems to collaborate, yet collaboration is only safe when the organisation can constrain who or what may act across those channels. The practitioner takeaway is to pair interoperability strategy with permission governance and auditability from the start.
Agentic adoption is exposing the limits of vendor-centric control models: The article rightly warns against lock-in, but the deeper issue is governance continuity. If identity and security depend on a single platform behaving perfectly, teams lose optionality and resilience when the AI environment changes quickly. Practitioners should expect more composable identity stacks, more protocol-aware control points and less tolerance for brittle, monolithic administration.
From our research library:
- Only 5.7% of organisations have full visibility into their service accounts, according to the Ultimate Guide to NHIs.
- Read next: AI Agent Identity Security Buyer's Guide
What this signals
Legacy infrastructure is becoming an identity problem first and a technology problem second: When AI adoption happens department by department, the organisation loses the ability to connect access, device trust and tool behaviour into one governance model. That is the point where security teams should expect blind spots to form around AI workflows rather than around the AI model itself.
Identity blast radius debt: This is the accumulation of every fragmented approval path, isolated permission set and disconnected audit trail that makes an automated tool harder to contain once it is live. The practical signal is that agentic adoption needs governance designed for change velocity, not annual review cycles.
The right response is to move from control-by-exception to control-by-design, with identity and device trust managed together and with explicit governance for MCP and A2A-connected workflows.
For practitioners
- Consolidate identity and device governance Build a shared control plane for users, devices and automated tools so permissions and trust state are visible in one place.
- Review departmental AI tool access Inventory locally adopted AI tools, then map which identities can access data, systems and delegated actions outside central approval flows.
- Align interoperability with authorisation Apply explicit permission checks and logging to MCP and A2A-connected workflows so collaboration does not outpace governance.
- Reduce legacy workflow friction Replace manual approval chains and fragmented data handling with managed identity workflows that can be audited and changed quickly.
- Build exit options for critical capabilities Qualify more than one vendor for high-impact identity and AI control points so platform changes do not stall governance response.
Key takeaways
- Legacy infrastructure weakens agentic AI governance because it splits identity, device and permission control across systems that do not move at the speed of AI adoption.
- More than 50% of enterprises say legacy IT slows scaling, which shows that technical debt is now a direct governance constraint rather than a background maintenance issue.
- Security teams should treat departmental AI adoption as an access governance event, not just a tooling decision, and bring it under unified identity control.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Agentic AI Top 10 addresses the attack and risk surface, while NIST AI RMF, NIST Zero Trust (SP 800-207) and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | AI tools acting with delegated access can abuse identity scope when legacy controls lag. |
| ASI02 — Tool Misuse | The article centers on AI tools being used beyond central visibility and intended governance. | |
| Recommendation — Map delegated AI access to ASI03 and constrain privilege scope before tools spread across departments. Use ASI02 to review how AI tools are selected, approved and monitored in departmental deployments. | ||
| NIST AI RMF | GOVERN — AI Governance and Accountability | The article is fundamentally about governance structures that can keep pace with AI adoption. |
| Recommendation — Establish AI governance ownership, approval and oversight under GOVERN before scaling tool adoption. | ||
| NIST Zero Trust (SP 800-207) | Zero Trust principles — Zero Trust principles | The article explicitly frames the response in Zero Trust terms for identity and device access. |
| Recommendation — Apply Zero Trust principles to verify users, devices and tools continuously as AI usage expands. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Unified permission control is central to the article's argument about agentic governance. |
| Recommendation — Use PR.AA-05 to review and limit permissions for humans and automated tools in one model. | ||
Key terms
- Identity Blast Radius: The amount of damage a compromised identity can cause across systems, data, and infrastructure. In NHI environments, it is shaped by permissions, network reach, and administrative capability rather than by the credential alone. Reducing blast radius is a containment strategy that limits lateral movement and data exposure.
- Agentic AI: Autonomous AI systems capable of planning, deciding, and taking actions, including calling APIs, writing code, and orchestrating other agents, with minimal human oversight. Agentic AI introduces new NHI risks as agents must authenticate to external services.
- Unified Control Plane: A unified control plane is an identity architecture where discovery, access governance, audit, and response operate across humans, machines, and AI agents together. It reduces blind spots caused by siloed tooling and gives security teams context for decisions about permissions, data, and containment.
- Interoperability Risk: Interoperability risk is the security exposure created when separate platforms, providers, or environments must connect and exchange data or identities. In metaverse settings, those connections can expand the attack surface, create inconsistent controls, and expose gaps in authentication, authorization, privacy, and incident response across different services.
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 9, 2026.
Updated on October 8, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org