TL;DR: AI-related attacks have increased nearly 490 percent year over year, according to Grip Security, but the deeper problem is that most AI governance tools stop at discovery and do not enforce control across SaaS identities, OAuth links, and non-human access. Governance without enforcement is observation, not security.
At a glance
What this is: This is an independent analysis of AI governance tools, with the central finding that most platforms still detect risk better than they control it.
Why it matters: It matters because AI risk in enterprises now flows through SaaS, OAuth, and non-human identities, which means IAM, IGA, and security teams need enforcement, not just visibility.
By the numbers:
- AI-related attacks have increased nearly 490 percent year over year.
- 80 percent of AI-related incidents involve sensitive or, tive or regulated data.
👉 Read Grip Security's webinar on the best AI governance tools for enterprises
Context
AI governance tools have become crowded because enterprises now have to manage AI usage, AI risk, and access paths at the same time. The core problem is that many products are still built to discover AI presence rather than enforce who and what can use it, which leaves identity governance exposed across SaaS environments and OAuth-connected integrations.
In practice, that means security teams cannot treat AI governance as a periodic assessment exercise. Once AI is embedded across thousands of SaaS applications, the control point shifts to identity, access, and integration governance, including non-human identities that can act on behalf of users or services without direct human oversight.
Key questions
Q: How should security teams govern AI features embedded in SaaS applications?
A: Treat embedded AI as a machine identity problem with data access implications. Inventory the feature, map the connected permissions, define what data it may use, and monitor retention and sharing paths. If the AI feature can read corporate content, it needs explicit approval, logging, and periodic review like any other privileged integration.
Q: Why do discovery-only AI governance tools leave material risk behind?
A: Because discovery tells you where AI exists, but not whether it is still appropriately authorised or constrained. If a tool cannot enforce scopes, revoke stale access, or limit connected identities, it records risk without reducing it.
Q: What do security teams get wrong about Shadow AI?
A: They often treat Shadow AI as an approval problem for software, when it is usually also an identity problem. The hidden risk can be an undocumented token, an over-permissioned service account, or an autonomous agent with unreviewed reach. Inventory the identity layer before you decide the tool is the issue.
Q: Should AI governance sit with security, IAM or the business?
A: It should be shared, with security setting control requirements, IAM managing identities and permissions, and the business owning approved use cases and risk acceptance. AI governance fails when it is treated as a side project because the controls span multiple teams. Clear ownership, access review and change control are the minimum coordination points.
Technical breakdown
Why AI governance breaks at the identity layer
Most AI governance tools are built around discovery, classification, or policy recommendation. Those functions help teams understand exposure, but they do not stop an identity from accessing data or a connected application from expanding the blast radius. In SaaS environments, AI risk is rarely isolated to the model itself. It is created by the combination of OAuth grants, delegated access, and non-human identities that can move data between systems. Governance that does not reach the identity layer cannot enforce least privilege or constrain what AI-connected tools can touch.
Practical implication: map AI governance controls to identity and integration points, not just to model inventories.
OAuth-connected SaaS creates the real control plane
OAuth is central because it turns application trust into delegated access. Once an AI-enabled SaaS app is authorised, the security problem becomes how much access it inherited, what data it can reach, and whether that access is still appropriate as the environment changes. Traditional posture tools can report the connection, but they often cannot continuously limit the scope of the delegated permissions. That is why AI governance in SaaS is fundamentally an access governance problem, not only a detection problem.
Practical implication: review OAuth grants as living entitlements and apply continuous access enforcement to them.
Shadow AI expands through non-human identities
Shadow AI is not just unsanctioned model use. It also includes unmanaged AI-enabled services, bots, and integrations that inherit permissions outside formal review. These non-human identities can create risk faster than human access processes can recertify it, especially when ownership is unclear and enforcement is weak. The category gap is that many governance programs still assume AI can be managed like a discrete application, when in reality the exposure is distributed across service accounts, tokens, and connected workflows.
Practical implication: include non-human identities in AI governance inventories, reviews, and offboarding workflows.
NHI Mgmt Group analysis
The AI governance market is solving visibility faster than it is solving control. That matters because discovery without enforcement creates a false sense of coverage. In SaaS environments, AI risk is shaped by delegated access and non-human identities, so governance that stops at inventory will miss the actual exposure path. Practitioners should treat this as a control-plane problem, not a reporting problem.
Identity and access governance now defines whether AI governance is real or cosmetic. The category boundary is shifting from model oversight to access enforcement across users, service accounts, tokens, and OAuth-linked apps. That shift means AI governance cannot sit outside IAM and IGA. Security teams should expect the strongest controls to be the ones that can govern permissions continuously, not the ones that merely classify AI usage.
Shadow AI is becoming an identity sprawl problem, not just an adoption problem. When AI is embedded across thousands of SaaS applications, ownership breaks down and enforcement becomes fragmented. The governance challenge is no longer just finding AI, but understanding which identities can act through it and whether those identities can be revoked, scoped, or reviewed. Practitioners should align AI governance with non-human identity lifecycle management.
Continuous control is now the category requirement, not an advanced feature. Periodic assessment cannot keep pace with SaaS-native AI use because permissions, integrations, and usage patterns change continuously. That makes continuous enforcement the relevant dividing line between mature and immature programs. Security leaders should re-baseline their AI governance strategy around live access control rather than static review cycles.
From our research:
- 85% of organisations lack full visibility into third-party vendors connected via OAuth apps, according to The State of Non-Human Identity Security.
- Only 1.5 out of 10 organisations are highly confident in their ability to secure NHIs, according to The State of Non-Human Identity Security.
- For lifecycle context, see Ultimate Guide to NHIs , Lifecycle Processes for Managing NHIs for how provisioning, rotation, and offboarding should be governed.
What this signals
AI governance will increasingly be evaluated as an access-control discipline, not a policy-writing exercise. As SaaS-native AI spreads, the teams that win are the ones that can enforce scope, ownership, and revocation across identities, integrations, and delegated access paths. Discovery tools will still matter, but they will no longer be sufficient on their own.
With 85% of organisations lacking full visibility into third-party vendors connected via OAuth apps, the control gap is already well established. The next programme milestone is not more inventory, but tighter enforcement around who can authorise, inherit, and retain access through AI-connected SaaS services.
For practitioners
- Inventory AI access at the identity layer Map every AI-enabled SaaS application to the users, service accounts, tokens, and OAuth grants that can reach it. Include unmanaged and shadow usage, not only sanctioned platforms.
- Treat OAuth permissions as governed entitlements Review delegated access continuously, not during annual audits. Revoke stale grants, reduce scopes where possible, and tie approval to business ownership.
- Bring non-human identities into AI governance Extend recertification, offboarding, and ownership assignment to bots, API tokens, and service accounts that interact with AI-enabled SaaS tools.
- Separate discovery from enforcement in tool selection Require any AI governance platform to show how it constrains access in real time, not only how it identifies risky usage or produces reports.
Key takeaways
- AI governance fails when it stops at discovery and does not enforce access across identities and SaaS integrations.
- The evidence points to a category-wide visibility and control gap, with OAuth-connected third-party access still widely unmanaged.
- Security teams should align AI governance with IAM, IGA, and NHI lifecycle controls if they want continuous enforcement rather than periodic 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 address the attack and risk surface, while NIST CSF 2.0, NIST Zero Trust (SP 800-207), NIST SP 800-53 Rev 5 and NIST AI RMF set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-03 | The article centers on non-human access governance and OAuth-connected identity risk. |
| NIST CSF 2.0 | PR.AC-4 | The post is about continuous access control across identities and integrations. |
| NIST Zero Trust (SP 800-207) | The control model depends on continuous verification of access paths. | |
| NIST SP 800-53 Rev 5 | IA-5 | OAuth tokens and delegated credentials are authentication artifacts requiring lifecycle control. |
| NIST AI RMF | GOVERN | AI governance requires accountability, ownership, and ongoing oversight. |
Map AI-connected SaaS access to NHI-03 and enforce scoped, reviewable, and revocable entitlements.
Key terms
- 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.
- OAuth Grant: An OAuth grant is the delegated permission an application receives to act on a user's behalf without storing the user's password. In NHI governance, it should be treated as a standing identity relationship with scope, ownership, and revocation requirements, not as a one-time setup detail.
- 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.
- Non-Human Identity (NHI): A digital identity assigned to a non-human entity such as a software application, service account, API key, bot, machine, or AI agent that enables it to authenticate and interact with systems without direct human involvement. NHIs now outnumber human identities in most enterprises by 25 to 50 times.
What's in the full article
Grip Security's full webinar covers the operational detail this post intentionally leaves for the source:
- Side-by-side comparison of AI governance tool categories and where each one stops short on enforcement
- Practical breakdown of SaaS identity and OAuth control points that create AI exposure
- Vendor-specific examples of how teams can monitor and constrain non-human access in live environments
- Implementation guidance for shifting from discovery-led governance to continuous control
Deepen your knowledge
NHI governance, agentic AI identity, and machine identity security 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, IGA, or security programme, it is worth exploring.
Published by the NHIMG editorial team on August 15, 2026.
NHI Mgmt Group — the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org
