TL;DR: Enterprise AI adoption is outpacing governance maturity, with Delinea arguing that visibility, accountability, and training must extend across AI systems, shadow AI, and AI agents. The deeper issue is that organisations are trying to govern a fast-moving identity layer with controls designed for slower, centrally managed programmes.
At a glance
What this is: A Delinea blog argues that enterprise AI governance is lagging adoption and that visibility, accountability and training must extend across embedded AI, shadow AI and AI agents.
Why it matters: IAM, risk and security teams need to treat AI governance as an identity and lifecycle problem, because unmanaged AI usage creates access, oversight and policy gaps across the enterprise.
Context
Enterprise AI governance is the gap between how quickly AI is being embedded into business workflows and how slowly organisations are adapting oversight, policy and control processes. In practical terms, the article treats AI usage as an enterprise identity governance issue, not just a model-risk or IT project.
The central problem is that many programmes still assume AI can be governed through static documentation and one-time approval, even as usage expands through shadow AI, AI agents and vendor-added features. The result is a control environment that cannot reliably inventory what exists, assign accountability or sustain review at the pace of adoption.
Key questions
Q: What breaks when AI governance is limited to policy documents and dashboards?
A: What breaks is enforcement. Policy documents can describe acceptable use, but they do not prove that access was approved, data was scoped correctly, model versions were controlled, or exceptions were remediated. Dashboards help visibility, but without workflow integration and immutable logs, they do not close the gap between intent and operational behaviour.
Q: Should organisations prioritise AI inventory discovery or policy writing first?
A: Inventory discovery should come first because you cannot govern what you cannot see. A policy without a reliable picture of which AI tools are in use will miss shadow AI, unmanaged access paths, and the endpoints that need enforcement most.
Q: How do teams know whether AI governance is actually working?
A: Look for evidence that every AI interaction can be traced end to end, from identity and intent to output and enforcement. If auditors can ask for a transaction and receive a complete record in hours, not weeks, the programme is producing usable control evidence rather than just documentation.
Q: Why do AI agents and shadow AI complicate governance in enterprise environments?
A: AI agents complicate governance because they can access data through multiple connectors, often outside managed approval paths. Shadow AI adds another blind spot when employees use tools on personal or unregistered accounts. That combination makes it hard to know what data is exposed, who approved access, and whether controls are actually operating as intended.
Technical breakdown
Why shadow AI breaks visibility-based governance
Shadow AI is unmanaged or unsanctioned AI use that sits outside the official inventory, approval and oversight process. In governance terms, the problem is not only that the tools exist, but that their data access, prompt handling and downstream outputs are often invisible to the teams responsible for risk and compliance. When AI is embedded into business units, products and third-party services, a central policy document cannot govern what it cannot discover. The control failure is at the inventory layer: if you cannot identify the AI surface area, you cannot classify it, monitor it or tie it to accountable ownership.
Practical implication: build discovery and inventory processes before attempting policy enforcement.
Why AI governance needs lifecycle ownership, not just policy text
The article’s governance model treats responsible AI use as an ongoing lifecycle responsibility, not a one-time approval exercise. That means defining who owns review, approval, monitoring and auditing for AI systems as they change over time. This is the same structural logic used in mature IAM and IGA programmes: policy sets intent, but lifecycle controls enforce accountability across introduction, change and retirement. Without named ownership for the full AI lifecycle, governance degrades into statements of principle that are difficult to operationalise or audit.
Practical implication: assign explicit lifecycle owners for each AI system and agent, including monitoring and retirement responsibilities.
How training turns AI policy into enforceable behaviour
The article is clear that policy without awareness fails. Training matters because AI governance has a human decision layer even when the technology is automated, especially where employees, partners and customers can unintentionally create risk through unapproved use. The operational lesson is to embed AI into established awareness programmes rather than treating it as a separate campaign. That approach makes governance repeatable, auditable and aligned to broader conduct and security expectations. It also helps close the gap between what a policy says and what people actually do with AI tools in day-to-day work.
Practical implication: fold AI use cases into annual awareness, acceptable-use and risk training.
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
Enterprise AI governance is becoming an identity governance problem, not just a model-risk problem. The article shows that AI now sits inside workflows, products and decision paths, which means governance has to track who or what can act, approve and expose data. That is familiar territory for IAM and IGA teams, but the subject is now AI systems and agents rather than only people. The practical conclusion is that AI governance needs the same ownership discipline applied to every other high-impact identity surface.
Shadow AI is the clearest sign that discovery now matters more than policy volume. A programme cannot govern an AI estate it cannot see, and that is why inventory and classification are the first meaningful controls. This is where many organisations confuse documentation with governance. The field should treat undiscovered AI usage as a standing governance defect, because every hidden instance creates unmanaged access and unreviewed risk.
Responsible AI use needs a lifecycle model because accountability has to survive change. The article’s emphasis on review, approval, monitoring and auditing points to a governance pattern that extends beyond initial deployment. That aligns with core lifecycle disciplines in identity security, where ownership must continue after go-live. The practitioner lesson is that AI oversight fails when it stops at policy sign-off instead of following the system through modification and retirement.
Training is the control that turns governance from intent into repeatable behaviour. The article is right to place awareness alongside policy, because users will route around controls when AI is easier to use than the approved path. That makes education part of the control stack, not a communications afterthought. The broader implication is that AI governance programmes will be judged by whether they change day-to-day behaviour, not whether they publish a principles statement.
Shadow AI governance debt is now a named risk category. The useful concept here is the gap between unsanctioned AI use and the organisation’s ability to inventory, review and own it. That debt accumulates quickly because AI can appear in business units, vendor products and embedded workflows without a clean handoff to security or risk teams. Practitioners should treat that debt as a measurable programme weakness, not an abstract concern.
From our research library:
- 52% of respondents see AI security decision-making power shifting toward platform and infrastructure teams rather than the executive suite, according to the 2026 Infrastructure Identity Survey.
- Only 44% of organisations have implemented any policies to manage their AI agents, despite 92% agreeing that governing AI agents is critical to enterprise security, according to the 2026 Infrastructure Identity Survey.
- Read next: Agentic AI Identity Risk Board Briefing
What this signals
Shadow AI governance debt: organisations that cannot inventory AI usage, ownership and review paths accumulate unmanaged risk faster than policy teams can reduce it. The control failure is visibility first, because hidden AI is also unreviewed AI.
The 44% of organisations that have implemented any policies to manage their AI agents show how far policy and operational control can diverge, according to the 2026 Infrastructure Identity Survey. That gap matters because governance only works when discovery, ownership and enforcement move together.
For practitioners
- Map every AI entry point Inventory employee-used tools, embedded AI features, vendor-added AI functions and AI agents so governance starts from what actually exists, not what policy assumes exists.
- Assign lifecycle ownership Name a business and security owner for each AI system or agent who is responsible for approval, monitoring, audit evidence and retirement decisions.
- Classify AI use by risk Create a risk classification model that distinguishes low-impact experimentation from systems that influence customer interaction, operations or decision-making.
- Embed AI into existing training Add AI use cases, approved tool boundaries and reporting expectations into annual awareness, acceptable-use and partner training programmes.
- Establish cross-functional oversight Use a governance committee or PMO-style model to coordinate legal, IT, operations, product and security review for material AI deployments.
Key takeaways
- Enterprise AI is being adopted faster than most governance programmes can inventory, classify and control.
- The strongest failure mode is not a missing principle statement, but a missing operational model for visibility and lifecycle ownership.
- IAM, security and risk teams should treat AI governance as an ongoing control discipline that must be embedded into discovery, review and training.
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 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 governance gaps become identity and privilege problems when agents operate with unclear authority. |
| ASI10 — Rogue Agents | Shadow AI maps directly to unmanaged or undeclared agent behaviour outside governance oversight. | |
| Recommendation — Map AI agent authority to ASI03 and restrict the actions each agent can take without review. Inventory and contain undeclared AI agents under ASI10 before they spread into business workflows. | ||
| NIST AI RMF | GOVERN — AI Governance and Accountability | The article focuses on governance structures, ownership and accountability for enterprise AI. |
| MAP — AI Risk Mapping | Visibility into AI usage and risk classification is central to the article's governance approach. | |
| Recommendation — Use GOVERN to assign decision rights, oversight roles and accountability for AI systems and agents. Apply MAP to inventory AI use cases, classify risk and document where oversight is currently missing. | ||
| NIST CSF 2.0 | GV.OC-01 — Organisational Context | The article frames AI as an enterprise-wide governance issue spanning business functions. |
| GV.RM-01 — Risk Management Strategy | The post argues that AI governance must be embedded into enterprise risk management. | |
| PR.AA-05 — Access Permissions, Entitlements and Authorizations | AI systems and agents need controlled permissions if they are to be governed operationally. | |
| Recommendation — Define how AI use fits the organisation's mission, workflows and governance boundaries. Embed AI risk decisions into the enterprise risk strategy and review them on a recurring basis. Limit AI permissions to the minimum entitlements needed for each approved use case. | ||
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 Governance: AI governance is the set of controls used to discover, classify, approve, restrict, monitor, and revoke AI-enabled access. It connects identity, data, and policy so organisations can manage what AI can reach, what it can share, and when it should be stopped.
- AI Lifecycle: The AI lifecycle is the end-to-end path from problem framing to retirement. It covers the decisions that shape a system’s purpose, data, behaviour, deployment, oversight, and decommissioning. In practice, it is the governance map that shows where risk enters and where accountability must stay active.
- Runtime AI Governance: Runtime AI governance is control applied while the interaction is happening, rather than before deployment or after an incident. It combines discovery, policy enforcement, output inspection, and audit logging so that AI use can be managed in live enterprise conditions.
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 or identity security 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