TL;DR: C1.ai details how enterprise AI moved from chatbots and copilots to custom agents that let business teams build real workflows, while bans push users toward shadow AI on personal machines. Blocking AI pushes usage onto personal machines and ungoverned paths, so identity control shifts from prevention to sanctioned sign-in and authorization.
At a glance
What this is: NHI Mgmt Group's analysis, drawing on C1.ai's blog post: Enterprise AI adoption is shifting from experiments to governed daily work, and the key failure mode is shadow AI created by blocking rather than enabling builders.
Why it matters: IAM, IGA, and security teams now need to treat AI builders as identity subjects, because the control problem is who can sign in, authorize, and publish agents, not whether employees will build them.
👉 Read C1.ai's analysis of governed AI builders and shadow AI
Context
Enterprise AI changes the identity surface when building moves out of a small engineering group and into sales, marketing, finance, and operations. The security question is no longer whether people will experiment with AI, but whether the organisation gives those builders a governed way to authenticate, authorize, and share what they create.
When policy blocks the approved path, users often route around it and keep working. That creates shadow AI: unmanaged apps and agents that can carry company data outside the controls meant to govern them.
Key questions
Q: How do you govern AI builders with sign-in and authorization?
A: Give builders a controlled platform where identity, authorization, and publishing are built in rather than bolted on. The key is to make access decisions explicit at creation time, then tie every app or agent to an accountable owner, a defined scope, and a revocation path when it is no longer needed.
Q: Why does banning employee AI use create shadow AI?
A: Banning employee AI use creates shadow AI because people still use personal devices and unmanaged services to get work done, but outside approved identity, logging and data controls. That bypasses account governance, approved sign-in, prompt review, retention rules and revocation. A safer model is to approve specific tools, assign accountable owners, and enforce SSO, DLP and usage policies.
Q: How should teams inventory AI agents that are built outside central IT?
A: Teams should record every externally built AI agent as a governed identity subject with an owner, intended purpose, tool list, data sources, model version, and approval path. Use a central register, unique agent ID, scoped credentials, and periodic review so shadow agents can be discovered, constrained, and revoked when no longer needed.
Q: What is the difference between an AI agent and a chatbot for security purposes?
A: A chatbot mainly responds inside a constrained conversational flow, while an AI agent can plan tasks, call tools, and execute actions across systems. For security teams, that difference matters because a chatbot is usually a user interface, but an agent is an identity-bearing actor with potential authority. The governance model must match the higher risk.
Technical breakdown
Why sanctioned AI needs sign-in and authorization
Once non-engineering staff start building apps and agents, identity becomes part of the runtime path. A builder who can create useful automation still needs a way to prove who they are, what they may publish, and which data or actions the agent can access. Without that layer, the organisation gets speed without governance, and every app becomes a separate trust decision. In practice, the control boundary moves from the model itself to the identity and authorization layer around the model. Read the Ultimate Guide to NHIs: Standards.
Practical implication: Treat sign-in and authorization as mandatory platform primitives for any internal AI builder environment.
Why blocking AI produces shadow AI
A blanket ban does not remove demand for automation or productivity gains. It changes where the work happens, often onto personal devices and unsanctioned services that the organisation cannot inventory, govern, or revoke. That makes shadow AI an identity problem as much as a tooling problem, because the organisation loses control over who is building, where access resides, and which data flows leave the managed environment.
Practical implication: Design policy around controlled access paths instead of assuming prohibition will suppress usage.
How the builder model expands the attack surface
The agentic enterprise distributes creation rights widely, which means access is no longer concentrated in a small platform team. Every new builder can introduce new identities, tokens, shares, and data connections, often faster than central IT can review them. That changes the security model from one of a few privileged deployment points to many distributed trust decisions. The result is less about one large application risk and more about a growing identity mesh that needs lifecycle control.
Practical implication: Inventory who can create, deploy, and share agents before the builder population outgrows your governance process.
Breaches seen in the wild
- McKinsey AI platform breach: McKinsey AI platform hack exposed 46M chats and sensitive data.
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
Shadow AI is the predictable outcome of prohibition-first governance. When employees are told they cannot build AI where they work, they do not stop building. They shift the activity to personal devices and ungoverned services, which removes inventory, policy enforcement, and revocation from the organisation's control. The implication is that AI governance fails first as a distribution problem, not a model problem.
The new control point is sanctioned identity, not model access. The article's real signal is that builders now sit across the business, so the relevant question is who can sign in, who can authorize an app, and who can publish it for others to use. That is an identity governance question wrapped around AI behaviour, and it needs the same lifecycle discipline used for other non-human access paths.
AI builders turn application governance into a delegated identity problem. The more employees can create apps and agents on their own, the more the platform has to govern delegated authority, shared access, and offboarding of what they create. Machine identity vs workload identity vs NHI helps separate the terms, but the practical issue is simpler: every builder-generated agent becomes a governed identity subject the moment it can act on data or systems.
Discovery becomes part of governance once builders proliferate. The article makes clear that some apps already exist before the sanctioned path does. That means governance is not only about approving new builders, but also finding what was created outside the approved path and deciding whether it can be brought under control. Practitioners should treat discovery and policy as a single programme, not two separate tasks.
Read next: Ultimate Guide to NHIs — Standards
What this signals
Governed AI needs an identity layer before it needs more experimentation. The article shows that the real differentiator is not whether employees can prompt models, but whether they can sign in, authorize apps, and share what they build inside a controlled environment. That pushes AI governance into the same operational space as IAM and lifecycle management, where ownership and revocation matter as much as innovation.
Shadow AI will keep expanding wherever policy is easier to bypass than to follow. Organisations should expect builders to use whatever path gets the job done, which means discovery, approval, and access control have to work together. The practical test is whether the secure route is easier than the workaround, because if it is not, the workaround becomes the default.
For practitioners
- Establish a sanctioned AI builder path Provide a governed place for employees to build, deploy, and share apps and agents instead of forcing them into unsanctioned tools.
- Require sign-in and authorization for internal agents Make identity, access approval, and app publishing controls mandatory before any employee-built agent can interact with company data or workflows.
- Map where shadow AI can emerge Look for the personal devices, unapproved services, and ad hoc workarounds employees will use when policy blocks the approved route.
- Inventory builder-created apps and agents Track who created each app, what it can access, and whether it still has an accountable owner for ongoing review and removal.
Key takeaways
- Enterprise AI becomes an identity governance issue when building shifts from experiments to business users who need sanctioned access, publishing, and oversight.
- Blocking AI rarely removes usage; it pushes builders into shadow AI paths the organisation cannot inventory or revoke.
- Governance has to make the approved path easier than the workaround, or AI usage will grow outside the control plane.
Related NHI Mgmt Group resources: Ultimate Guide to NHIs: Standards
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, OWASP Non-Human Identity 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 Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Employee-built agents need governed identities and scoped authority, which this article makes central. |
| Recommendation — Bind agent creation, sign-in, and publishing to explicit identity controls and scope limits. | ||
| OWASP Non-Human Identity Top 10 | NHI-04 — Insecure Authentication | Sanctioned AI builders need controlled sign-in before they can deploy or share apps and agents. |
| NHI-05 — Overprivileged NHI | Custom agents can quickly accumulate access beyond what their business role needs. | |
| Recommendation — Enforce strong authentication for every builder and every agent that acts on enterprise systems. Review agent permissions against intended tasks and remove access that exceeds business need. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article centres on who can create, authorize, and publish AI apps inside the enterprise. |
| Recommendation — Define and enforce permissions for builder platforms, app publishing, and shared agent access. | ||
| MITRE ATT&CK | TA0006; TA0008 — Credential Access; Lateral Movement | Shadow AI and unmanaged agents can become pathways for credential exposure and broader reach. |
| Recommendation — Monitor for unmanaged agent activity that expands credential exposure or reaches new systems. | ||
Key terms
- Shadow AI: AI agents, copilots, or connected tools operating without full visibility or governance from security teams. Shadow AI becomes an identity problem when those systems authenticate with unmanaged tokens, service accounts, or OAuth apps that can reach production resources.
- AI Agents: AI agents are autonomous software entities that act within organisational environments and make runtime decisions within assigned boundaries. They can hold identities, authenticate to systems, and exercise permissions, which makes them comparable to other non-human identities that require inventory, governance, and continuous activity monitoring.
- Authorization: Authorization is the decision about what an authenticated identity is allowed to do. In NHI and IAM practice, it covers scope, duration, and allowable actions, and it is the layer that most directly controls blast radius when access is active.
- Identity Governance: Identity governance is the set of controls that defines who approves access, who owns it, how it is reviewed, and when it is removed. In practice, it turns identity management from a deployment task into a durable control system that can withstand audits, organisational change, and operational growth.
What's in the full article
C1.ai's full blog post covers the operational detail this post intentionally leaves for the source:
- How the platform frames sign-in and authorization for employee-built apps and agents
- The app marketplace and deployment model used to keep builders inside a governed path
- Why discovery of previously built shadow AI matters before central governance can work
- The practical shift from banning AI use to controlling how builders create and share it
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 November 16, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org