Identity-data blind spots in the AI era expose access risk

At a glance

What this is: Cyera argues that fragmented identity records and disconnected data visibility leave security teams unable to see who can access sensitive data across environments.

Why it matters: That matters because IAM, NHI, and human identity programmes all fail when access cannot be traced back to a single accountable identity, especially as AI tools act on the same entitlements.

👉 Read Cyera and Okta's analysis of identity and data access blind spots in AI

Context

Identity and data security break down when organisations treat the same person as multiple separate accounts across SaaS and cloud. The core problem is not just where sensitive data lives, but whether security teams can reliably map each identity to every dataset it can touch.

That gap becomes more serious when copilots and other AI tools are layered onto existing access models. If identity consolidation is incomplete, AI can inherit fragmented permissions, which makes overexposure harder to detect and harder to limit.

Key questions

Q: How should security teams control sensitive data access when identities are fragmented across multiple systems?

A: Security teams should build a single access view that ties each person to every account, dataset, and authentication state across the environment. That lets IAM and data security teams evaluate least privilege consistently instead of relying on separate reports from each platform. Without that consolidation, governance decisions will always be partial.

Q: Why do fragmented identities make AI access risk harder to govern?

A: Fragmented identities make AI risk harder to govern because assistants and copilots can only be limited as well as the user accounts they inherit. If one person has inconsistent permissions across systems, AI can surface or act on data from the most exposed account. That makes identity normalisation a prerequisite for safe AI adoption.

Q: What do security teams get wrong about least privilege in SaaS and cloud environments?

A: Teams often treat least privilege as a role design exercise when the real problem is entitlement drift across multiple identities. A user may look compliant in one platform and over-permissioned in another. Effective least privilege requires cross-platform recertification, not isolated clean-up.

Q: How can organisations tell whether MFA enforcement is actually consistent across identities?

A: They need to verify MFA at the account level, not assume it applies to the person as a whole. A user with three accounts and one weak login path still has an exposed identity. Consistency means every active account tied to that user is protected, especially the account with the broadest data reach.

Technical breakdown

Why fragmented identities create access blind spots

The technical failure here is identity fragmentation across platforms. A single employee may appear as separate accounts in Microsoft 365, Google Workspace, Snowflake, and cloud services, while each system records access independently. Without identity enrichment, security tools cannot reliably correlate those records into one access picture. That means entitlement review, MFA coverage, and data-access mapping all become partial views rather than governance controls. Practical implication: build a consolidated identity layer that normalises user accounts before you attempt least-privilege reviews or data-access enforcement.

Practical implication: normalise accounts into one identity record before entitlement review or access enforcement.

How identity enrichment changes data access governance

Identity enrichment links a known user to the sensitive data that same user can reach across environments. In practice, that turns identity from a directory object into an access context signal, which is essential for policy decisions that depend on role, dataset sensitivity, and authentication strength. This is especially useful when access is spread across SaaS and cloud platforms, because the question is no longer only who the user is, but what that user can actually retrieve or expose. Practical implication: tie identity records to data classifications so access decisions can be evaluated consistently across systems.

Practical implication: tie identity records to data classifications so access decisions can be evaluated consistently across systems.

Why AI tools intensify existing permission problems

AI copilots and assistants do not create the underlying permissions problem, but they make it more consequential. When a tool can query or act on whatever the user can access, fragmented identity records translate into fragmented AI guardrails. If a finance user has excessive permissions in one system, an AI workflow can inherit that exposure and surface data the user should not be using in that context. Practical implication: treat AI enablement as an access-control review trigger, not just a productivity rollout.

Practical implication: treat AI enablement as an access-control review trigger, not just a productivity rollout.

Breaches seen in the wild

• DeepSeek breach — DeepSeek breach exposed 1M+ log lines and sensitive secret keys.
• Cisco DevHub NHI breach — IntelBroker exploited exposed Cisco credentials, API tokens and keys in DevHub.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Identity fragmentation is the governance gap, not just an operational inconvenience. When one employee is represented as multiple identities across SaaS and cloud, least privilege becomes impossible to assess with confidence. Security teams are not just missing a report view, they are missing the authoritative identity-to-data relationship that governance depends on. The practitioner conclusion is simple: if identity is fragmented, access governance is already incomplete.

Data access is now an identity problem, not a separate control domain. The article points to a real shift in control ownership, where data security and IAM can no longer operate on parallel tracks. If teams cannot answer who can touch a sensitive dataset, they cannot enforce meaningful role boundaries, recertification, or insider-risk monitoring. The field needs access governance models that join identity, data classification, and authentication state in one decision path.

AI inherits the access model you already failed to rationalise. Copilots do not remove the need for identity control, they amplify every entitlement you leave unresolved. That means over-permissioned users become higher-risk data sources once AI tools can act on their behalf or surface their data at machine speed. The practitioner conclusion is to govern AI through the same identity and data boundaries already required for human access, not as a separate exception.

Unified visibility is the named concept this category now needs. The practical failure mode is the absence of a single access picture that connects user identity, MFA state, and sensitive data reach. Without unified visibility, teams cannot distinguish legitimate multi-account usage from excessive entitlement accumulation. The practitioner conclusion is that visibility must be evaluated as a governance control, not a reporting feature.

From our research:

• 97% of NHIs carry excessive privileges, increasing unauthorised access and broadening the attack surface, according to Ultimate Guide to NHIs.
• Only 5.7% of organisations have full visibility into their service accounts, which shows how often access governance starts from incomplete inventory rather than control.
• For a broader view of identity failure modes, see 52 NHI Breaches Analysis for real-world cases where hidden access and weak lifecycle control widened impact.

What this signals

Unified access visibility is becoming a baseline expectation for any programme that wants to govern identity and data together. If you cannot tie identity state to data reach, every AI rollout, insider-risk review, and access certification will inherit the same blind spot.

The next governance step is to treat data access as part of identity lifecycle management, not a downstream reporting concern. That means recertification, MFA review, and privilege cleanup need to operate on the same identity graph, not separate spreadsheets.

For practitioners

• Consolidate fragmented identities into one access record Normalize employee accounts across Microsoft 365, Google Workspace, Snowflake, and cloud platforms so each person has one reviewable identity profile. Use that profile to compare actual data reach against job function and approval state.
• Map identity to sensitive data before enabling AI use cases Require a validated identity-to-data map before copilots or assistants can query business datasets. If the mapping is incomplete, pause rollout for those data domains until access boundaries are explicit.
• Review MFA coverage account by account, not user by user Check every account attached to the same employee for authentication gaps, because one protected login does not mean the whole identity is protected. Remediate the weakest account first, since that is the easiest path to account takeover.
• Re-certify excessive permissions against current role scope Use recertification to remove access that has accumulated across platforms over time. Focus on accounts that can reach financial, customer, or auditor-facing data but no longer match the user’s day-to-day responsibilities.

Key takeaways

• Fragmented accounts create a governance gap because security teams cannot reliably see who can access which sensitive datasets.
• The risk grows when AI tools inherit those same entitlements and can expose data faster than teams can review it.
• Programmes that unify identity, MFA state, and data access are better positioned to enforce least privilege across SaaS and cloud.

Key terms

• Identity Enrichment: Identity enrichment is the process of adding context to a basic account record so security teams can see who the user is, what systems they use, and what data they can reach. It turns fragmented login records into a governance object that can support access review and risk decisions.
• Access Visibility: Access visibility is the ability to see, in one place, which identities can reach which data, applications, and services. For IAM and data security teams, it is the difference between reviewing isolated permissions and understanding real blast radius across environments.
• Entitlement Drift: Entitlement drift is the gradual accumulation of permissions beyond what a user currently needs. It usually happens when role changes, project work, and temporary exceptions are never fully revoked, leaving the identity over-permissioned across one or more platforms.
• Identity-to-Data Mapping: Identity-to-data mapping connects a known identity to the sensitive datasets it can access. It is the governance layer that lets teams evaluate whether authentication, role scope, and data sensitivity are aligned instead of assuming the directory view is enough.

Deepen your knowledge

Identity-to-data governance is a core topic in our NHI Foundation Level course, the industry's only accredited NHI security programme. If your team is trying to unify access visibility across platforms, that course provides a practical starting point.

This post draws on content published by Cyera: Cyera and Okta: Eliminating Identity and Data Access Blind Spots in the AI Era. Read the original.