DSPM in 2026: Why shadow data and AI raise the stakes

At a glance

What this is: This is an analysis of how shadow data, AI adoption, and tool fragmentation are changing the DSPM baseline, with Securiti arguing that discovery and governance now have to extend across cloud, SaaS, and AI systems.

Why it matters: It matters because identity, access, and data teams increasingly have to govern not just data location, but which humans, workloads, and AI systems can reach sensitive data and whether those paths remain visible.

By the numbers:

• The 2026 GigaOm DSPM Radar evaluated 25 vendors across technical capabilities, emerging features, and business criteria.
• Data breaches involving multiple environments cost an average $5.05 million.

Context

DSPM is the discipline for discovering sensitive data, classifying it, and governing exposure across storage, analytics, and now AI pipelines. The article argues that the old model of isolating cloud, SaaS, and on-premises controls no longer matches how data moves, so incomplete visibility becomes a governance failure rather than just an inventory problem.

The identity angle is real because data access is now mediated by humans, service accounts, workloads, and AI agents that can ingest or transform information at machine speed. When classification and permissions lag behind that movement, organisations lose control over who or what can reach sensitive data, and that gap is now typical in complex enterprise estates.

Key questions

Q: How should teams govern shadow data that appears outside formal ownership?

A: Treat shadow data as an access governance problem, not only an inventory problem. The first step is to identify unowned repositories, assign accountable owners, and classify the data before allowing any downstream sharing or AI ingestion. If a repository cannot be tied to a business purpose and access model, it should be treated as an unmanaged exposure surface.

Q: Why does AI adoption increase the risk created by fragmented data controls?

A: AI systems compress the time between data discovery, movement, and reuse, so fragmented controls fail faster. When the same dataset is reachable through storage, copilots, and agent workflows, a weak permission model in one place can expose sensitive data across several others. That is why AI governance and data access governance now need to be aligned.

Q: How should security teams implement DSPM across multi-cloud and SaaS environments?

A: Start with API-based discovery across the platforms that hold regulated or business-critical data, then layer classification, access context, and monitoring on top. The key is consistency: the same policy logic should follow the data across cloud services, SaaS applications, and hybrid stores. Without that, visibility remains fragmented and exposure reports are incomplete.

Q: What happens when security tools for AI and data create new silos instead of integrating with existing workflows?

A: Adoption drops quickly. Teams treat the tool as shelfware because it adds friction without fitting into daily operations. The article’s message is that security findings must flow into existing systems such as IAM, GRC, DLP, and service workflows. When integration is missing, security loses both momentum and accountability.

Technical breakdown

Why shadow data breaks DSPM visibility

Shadow data is data created outside formal governance, often in ad hoc cloud repositories, project workspaces, or abandoned SaaS locations. The technical problem is not simply that the data exists. It is that discovery, classification, and ownership metadata never reach a consistent control plane, so security teams cannot reliably answer where the data lives, who can access it, or whether it is still needed. In practice, this creates ungoverned storage sprawl that resists cleanup and makes policy enforcement incomplete.

Practical implication: build discovery coverage for orphaned cloud and SaaS repositories before trying to tune classification policy.

How AI pipelines turn data governance into an access-control problem

AI systems do not just consume data, they create new data pathways. Training sets, retrieval layers, copilots, and AI agents all ingest content that may never have been approved for that context. Once sensitive data reaches those systems, access control is no longer only about storage permissions; it becomes about whether the AI workflow itself has lawful, scoped, and monitored access. That is why data governance and identity governance now intersect directly, especially where agents or models can move data without human review.

Practical implication: bind AI ingestion and retrieval to explicit access scopes, ownership, and reviewable policy controls.

Why siloed tools create measurable control gaps

Siloed data tools break the single-view requirement that modern DSPM depends on. Each point solution may detect risk in one environment, but the combined estate is what matters when data moves across cloud, on-premises, and AI systems. Fragmentation weakens lineage, duplicates effort, and leaves toxic combinations invisible because no single team sees the full path. The operational consequence is not just slower response. It is false confidence, where the organisation believes it has coverage but cannot prove control continuity across environments.

Practical implication: consolidate lineage, classification, and access intelligence into one governance view across data and AI assets.

Threat narrative

Attacker objective: The practical objective is to reach sensitive data through uncontrolled repositories or AI pathways and exploit that exposure for leakage, manipulation, or downstream breach impact.

1. Entry occurs through cloud expansion, SaaS proliferation, or AI adoption that creates new repositories and data flows faster than governance can track them.
2. Credentialed or privileged access to those repositories persists without a unified view of ownership, classification, or lifecycle control, so exposure accumulates in shadow data and AI pipelines.
3. Impact shows up as data leakage, regulatory exposure, model poisoning risk, and higher breach cost when fragmented tools fail to contain the full blast radius.

Breaches seen in the wild

• AI agent retail card theft campaign 2026: AI agents breached 27+ retailers for about $25 each, used cloud keys and a Secrets Manager dump, and stole 600,000+ payment cards.

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 data is now an identity governance problem as much as a storage problem. Once repositories can be created on demand and then forgotten, the control failure is not only incomplete discovery. It is the absence of a reliable ownership and access model for data that outlives the project that created it. Practitioners need to treat abandoned repositories as ungoverned access surfaces, not just stale assets.

AI access to data collapses the distance between data governance and NHI governance. When AI agents or model pipelines can ingest sensitive information, the organisation is governing non-human access paths, whether it labels them that way or not. That means the security question becomes who or what is authorised to move data into an AI system, and under what lifecycle constraints.

DataAI command models are becoming the practical answer to governance fragmentation. A single relationship graph across data, identities, policies, and AI systems is more than a visibility feature; it is the control model needed to prove whether access is still appropriate after data moves. Shadow data governance debt: the longer discovery, classification, and access decisions remain split across tools, the more likely the organisation is to accumulate hidden risk that no team can fully reconcile.

DSPM is maturing from a discovery product category into a board-relevant control plane. The article’s core message is that data security can no longer be assessed by asset counts or tool coverage alone. Practitioners should judge DSPM by whether it can preserve policy continuity across cloud, SaaS, on-premises, and AI workflows.

AI-era data security depends on proving safe data flow, not just finding data. Discovery without governed access becomes a catalog of exposure. The field is moving toward enforcement models that connect lineage, permissions, and AI usage so that data can be both found and constrained at the point of use.

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: Agentic AI Identity Guide

What this signals

Shadow data governance debt: the cost of delayed ownership, classification, and access decisions rises every time a team provisions a new repository and leaves it behind. The practical signal for security leaders is that discovery must be paired with lifecycle ownership, or the organisation will keep expanding its unmanaged surface faster than it can reduce it. according to the 2026 Infrastructure Identity Survey

AI systems are no longer passive consumers of information, they are becoming durable data access paths that need explicit governance. That shifts DSPM from a storage-centric discipline to a control problem that must account for who or what can reach data, why that access exists, and how quickly it can be revoked when the context changes.

For practitioners

• Expand discovery to shadow repositories Map cloud, SaaS, and on-premises locations that can be provisioned on demand, then verify which ones lack an owner, retention rule, or active classification.
• Tie AI ingestion to explicit access scopes Require approved datasets, policy checks, and monitored access paths before data can be consumed by copilots, internal models, or AI agents.
• Unify lineage and policy context Use a single governance view to connect data location, ownership, permissions, and downstream AI usage so fragmented tooling does not hide toxic combinations.
• Measure blast radius across environments Test whether a breach in one environment exposes data elsewhere, especially where cloud, analytics, and AI pipelines share incomplete or duplicated controls.

Key takeaways

• Shadow data, AI pipelines, and siloed tooling now create governance gaps that discovery alone cannot close.
• The exposure problem is no longer limited to where data sits, but to who and what can move it across systems.
• DSPM strategies that unify lineage, permissions, and AI context are better positioned to reduce hidden blast radius.

Key terms

• Shadow Data: Shadow data is sensitive information that exists outside the places security teams expect to find it. It often appears in testing copies, ad hoc exports, SaaS tools, or AI workflows, which makes it hard to govern with inventory-based controls alone.
• Data Security Posture Management: Data Security Posture Management, or DSPM, is the continuous discovery and monitoring of where sensitive data lives, how it is exposed, and where policy gaps exist. Its value rises when it feeds remediation rather than generating findings alone, especially in environments where AI expands the number of data paths.
• Data Lineage: The record of how data moves across systems, applications, and workflows. In security operations, lineage shows where sensitive data propagates, which identities touch it, and how a compromise could spread across connected environments.
• AI Data Governance: AI data governance is the set of rules, ownership decisions, and enforcement mechanisms that determine how data can be used by AI systems. It covers classification, access control, retention, and remediation, and it must account for both human users and autonomous software entities.

Deepen your knowledge

NHI Mgmt Group covers identity security, NHI governance, and agentic AI through independent research, practitioner guides, and the NHI Foundation Level course. It is designed for practitioners who need to connect access governance to the broader security programmes they run.