AI agent governance is outpacing enterprise IAM controls

At a glance

What this is: 1Password’s survey argues that AI adoption is outpacing enterprise governance, leaving visibility, enforcement, and lifecycle control gaps around AI tools and agents.

Why it matters: This matters because IAM, IGA, and PAM teams now have to decide how to govern AI-driven access that may never fit cleanly into existing approval, review, and revocation workflows.

By the numbers:

• Only 21% of security leaders say they have full visibility into AI tools used in their organisation.
• 54% of security leaders say their AI governance enforcement is weak.
• 56% of security leaders estimate that between 26% and 50% of their AI tools and agents are unmanaged.
• 1Password surveyed 200 North American security leaders.

Context

AI governance here means the policies, visibility, and enforcement controls that determine which tools can access corporate data and business systems, and under what conditions. The article argues that those controls are lagging behind the speed of employee AI adoption.

For identity teams, the challenge is not just tool sprawl. It is the fact that AI use now touches human IAM, NHI governance, and access governance at the same time, often outside formal onboarding, review, and offboarding processes.

Key questions

Q: What breaks when AI tools are adopted without formal identity review?

A: Without identity review, AI tools accumulate access outside normal procurement and offboarding controls, which makes it hard to track what data they can reach or who approved them. The result is unmanaged delegated access, hidden SaaS exposure, and token sprawl that persists after the original business need fades.

Q: Why do unapproved AI tools create more risk than traditional software use?

A: Unapproved AI tools can process user input in ways traditional applications do not, including storing prompts, retaining outputs, or reusing data to improve services. That creates exposure when employees paste regulated or sensitive information into systems the organisation does not oversee. The risk is not only misuse, but also loss of visibility into where data goes next.

Q: How should organisations govern AI agents that act as business units of work?

A: Organisations should govern AI agents as first-class non-human identities. That means assigning named ownership, defining the scope of permitted actions, separating duties such as propose versus publish, and requiring auditable logs and revocation paths before the agent is allowed into production. The governance model belongs in procurement, not just in the technical rollout.

Q: When do AI access controls fail in practice?

A: They fail when authorization happens after retrieval, when tool permissions are broad, or when response masking is treated as optional. In those cases the system has already pulled sensitive content into context or invoked a privileged action before policy intervenes. The safest design is source-side enforcement before the model sees the data.

Technical breakdown

Why AI tool visibility fails existing governance models

Visibility gaps appear when security teams cannot see which AI tools employees are using, how they are connected, or what data they touch. In practice, that means policy coverage and enforcement both become partial, because the control plane cannot govern what it cannot inventory. This is a governance problem before it is a tooling problem: if the approved estate is incomplete, recertification, blocking, and monitoring all start from bad data. The article’s point is that AI adoption makes this failure much more visible, not that it creates the failure from scratch.

Practical implication: build an AI usage inventory before trying to enforce policy at scale.

How AI access breaks traditional identity and access management assumptions

Traditional IAM assumes access is provisioned to a known subject, reviewed on a cycle, and removed when the relationship ends. AI agents and tools disrupt that model because access may be created ad hoc, borrowed from a human, embedded in a workflow, or left unmanaged outside formal lifecycle controls. That turns identity governance into a problem of delegated trust, not just account administration. The deeper issue is that access can be functionally real even when no formal identity record exists for the tool using it.

Practical implication: extend identity governance to AI-mediated access paths, not just named accounts.

Why unintentional data exposure is the biggest internal AI risk

The article frames the core issue as employees unknowingly giving AI tools access to sensitive data. That matters because many AI workflows are built around data ingestion, which can quietly widen the blast radius of a single user action. If a tool retains prompts, trains on uploaded content, or routes data into external services, data handling rules become inseparable from identity controls. This is where data governance and access governance converge: who can share what, with which tool, and under what retention terms.

Practical implication: define AI data-sharing rules as part of access governance, not as a separate policy memo.

Breaches seen in the wild

• 12,000 secrets in LLM training data: Truffle Security found 11,908 live API keys and passwords hard-coded in web pages captured by Common Crawl, a dataset used to train LLMs.
• DeepSeek database exposure 2025: An unauthenticated DeepSeek ClickHouse database exposed over a million log lines with plaintext chat history and API keys in 2025.

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

AI governance is now an identity governance problem, not a sidecar policy issue. The article shows that AI tools are entering enterprise workflows faster than visibility and enforcement can keep up. That pushes AI from a security awareness concern into the centre of IAM, IGA, and NHI oversight. Practitioners should treat AI access as part of the identity estate, because unmanaged use creates real governance blind spots.

Access-trust gap is the right name for what is breaking. The issue is not simply that employees are using unsanctioned AI. It is that organisations cannot reliably see, authorise, or constrain the access paths those tools create. Once AI can be connected through borrowed credentials, embedded integrations, or direct system links, traditional approval models stop describing reality. The implication is that governance must move from policy declaration to runtime visibility.

Traditional IAM assumes stable subjects and reviewable privilege, and that assumption is under strain. AI agents and tools can acquire access outside normal onboarding, persist without clear ownership, and interact with systems in ways that blur human and machine accountability. The result is a governance model that can record the existence of a policy but not necessarily the actual use of access. That is a lifecycle failure, not just a tooling gap.

Ephemeral AI governance records: the current programme model presumes that access is stable enough to inventory, certify, and revoke after the fact. That assumption weakens when access is assigned to tools and agents that appear, connect, and disappear across workflows faster than review cycles can observe. The implication is that practitioners must rethink governance timing, not just expand control coverage.

Security leaders should stop treating AI access as exceptional and start treating it as normal identity sprawl. The article’s four challenges are all variations of the same structural problem: access is moving outside the governance perimeter. For IAM and PAM teams, the practical consequence is that AI policies, audit trails, and lifecycle controls need to be designed as one operating model, not as separate point controls.

From our research library:

• 19% of organisations give AI systems dramatically more access than human employees, nearly one in five granting unrestricted privilege, according to the 2026 Infrastructure Identity Survey.
• Read next: AI Agent Identity Security Buyer's Guide

What this signals

AI governance is now a control-plane problem: the main question is no longer whether an organisation has an AI policy, but whether it can inventory tools, enforce access, and preserve evidence across the full lifecycle of AI-mediated access. That makes the AI usage inventory and enforcement model part of IAM architecture, not a separate awareness exercise.

The visible gap is structural, not cosmetic. When employees can attach AI tools to sensitive workflows without formal review, access governance loses the stable subject it needs to certify, revoke, and audit.

According to the 2026 Infrastructure Identity Survey, 19% of organisations give AI systems dramatically more access than human employees, nearly one in five granting unrestricted privilege. That pattern shows how quickly access assumptions collapse when AI is treated as just another productivity layer.

For practitioners

• Define an AI access inventory Map which AI tools, assistants, and agents are used by employees, which data they touch, and which business systems they connect to.
• Extend lifecycle governance to AI access Require ownership, review, recertification, and revocation paths for any AI tool or agent that can reach corporate data or systems.
• Separate approved from unsanctioned AI use Classify AI tools by business risk and decide which uses are monitored, which are blocked, and which require explicit approval.
• Set data-sharing rules for AI workflows Define what sensitive data may be shared with AI tools, then align those rules with user training and technical guardrails.
• Build audit evidence for AI-mediated access Capture who provisioned the access, what system it reached, and when it was reviewed so compliance teams can trace the full access path.

Key takeaways

• AI tools are now exposing a governance gap that sits at the intersection of human IAM, NHI oversight, and AI policy enforcement.
• The survey data shows a consistent control problem across visibility, enforcement, and lifecycle management, not an isolated configuration issue.
• Practitioners need to govern AI access as a first-class identity problem, with inventory, review, and revocation built into the operating model.

Key terms

• Access-trust gap: The gap between having a policy and actually enforcing it when access is requested. It appears when compliance, HR, or training data sits in separate systems from the access decision, allowing users to reach resources even though a required condition has not been met.
• AI-mediated access: Access that is initiated or transformed through an AI tool rather than a direct human action alone. In practice, the identity controls underneath the tool still determine who can reach data, what the tool can expose, and whether the resulting activity is auditable.
• Governed delegation: Governed delegation is the controlled assignment of access from a human or system owner to another actor under explicit rules, scope, and review. For AI tools and agents, it requires clear permission boundaries, traceable ownership, and a revocation path that is visible to identity teams.
• Unmanaged AI agent: An unmanaged AI agent is a deployed system that operates without being tracked in identity governance, secrets management, or privileged access controls. That means security teams cannot reliably see its owner, permissions, dependencies, or lifecycle state, making revocation and accountability difficult when risk changes.

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.