TL;DR: AI adoption is accelerating across GenAI SaaS, endpoint AI apps, and AI agents, while governance and visibility lag behind, creating uneven exposure patterns across organisations, according to Cyberhaven’s 2026 financial services report. The core problem is not AI use itself, but fragmented control over accounts, tools, and data paths.
At a glance
What this is: Cyberhaven’s financial services report says AI adoption is accelerating faster than governance, creating uneven exposure across GenAI SaaS, endpoint AI apps, and AI agents.
Why it matters: For IAM and security teams, the issue is not AI use itself but the loss of visibility and control once employees move across accounts, tools, and unmanaged environments.
Context
AI adoption in financial services now spans multiple control planes at once, including SaaS copilots, local AI apps, and agent-like workflows. That creates a governance gap because access, data movement, and user accountability no longer live in one place, so legacy visibility models miss how AI is actually being used.
The article’s core problem is split governance: the organisations moving fastest are also creating the hardest-to-observe data paths. For IAM, this matters because identity controls, account lifecycle decisions, and approval boundaries now influence AI risk as much as the tools themselves.
Key questions
Q: What breaks when AI adoption outpaces governance?
A: What breaks first is attribution. Teams lose visibility into which tools are in use, which data they can reach, and which actions were taken automatically versus manually. Once adoption is ahead of control design, security teams end up retrofitting policy around live workflows instead of governing them from the start.
Q: Why does fragmented AI infrastructure create security risk?
A: Fragmented AI infrastructure creates risk because each provider handoff can split responsibility for credentials, permissions, and logging. When teams stitch together GPU services, model hosting, and cloud platforms, the result is often inconsistent governance and hidden access paths. That makes it harder to prove who can reach what and under which conditions.
Q: What are the signs that AI governance controls are not keeping pace with adoption?
A: Common warning signs include unclear ownership for AI use cases, inconsistent approval processes, limited visibility into where sensitive data enters models, and weak evidence for audits or assessments. Teams also struggle when privacy, security, and legal review happen late or manually, because that usually means governance is reactive rather than embedded in the AI delivery process.
Q: How should teams govern AI agents before they scale?
A: Teams should require explicit ownership, narrowly defined scope, and auditable delegation for every agentic workflow. That means treating the agent as a governed identity-like actor rather than as a generic automation feature. Without that boundary, organisations will struggle to explain, review, or contain the actions the agent takes across systems.
Technical breakdown
Why fragmented AI usage breaks traditional visibility
AI usage fragments when employees can reach models through browser-based SaaS, installed desktop apps, personal accounts, and embedded workflows. Each path may use different identities, consent models, logging, and data handling rules, which means security teams no longer see a single, stable control surface. That is especially problematic in financial services, where regulated data often moves through ordinary productivity activity. The result is not simply more AI use, but more uncorrelated use that defeats inventory, review, and policy enforcement.
Practical implication: build inventory and control coverage around AI entry points, not around a single sanctioned platform.
How AI agents change the governance problem
AI agents are not just another application category. They can act across tools and accounts, which turns identity, delegation, and permission scope into live security issues rather than static policy questions. When an agent can request access, move data, and continue a task across systems, governance must cover the full decision chain, not only the user who initiated it. In practice, this pushes teams toward tighter control of delegated access, clearer ownership, and stronger logging around agent-mediated actions.
Practical implication: treat agent identity, delegated scope, and auditability as first-class governance controls.
Why data exposure rises as AI use spreads beyond IT oversight
The report links AI adoption to increasing sensitive data exposure because employees are using multiple tools and accounts, including personal and unmanaged environments. Once data is copied into those paths, standard DLP and IAM controls become less effective unless they extend across the actual workflow. This is a visibility and containment problem, not only a policy problem. The control gap is strongest where AI tools sit outside the organisation’s normal identity and endpoint boundaries.
Practical implication: extend monitoring and data control to unmanaged AI usage paths before exposure becomes normalised.
Threat narrative
Attacker objective: The practical attacker objective in this risk pattern is to exploit weakly governed AI usage paths to reach sensitive financial data or extend access beyond intended boundaries.
- Entry begins when employees interact with AI through GenAI SaaS apps, endpoint AI applications, personal accounts, or unmanaged environments outside the main IT boundary.
- Credentialed access and delegation then fragment across tools and sessions, making it harder to see who can move data, what context they inherit, and which actions are attributable to whom.
- Impact emerges as sensitive data flows through disconnected AI paths with reduced oversight, increasing the chance of exposure and weakening enforcement of corporate governance rules.
Breaches seen in the wild
- McKinsey AI platform breach: McKinsey AI platform hack exposed 46M chats and sensitive data.
- 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 adoption splitting faster than governance is the new enterprise risk pattern: the market is no longer dealing with uniform AI rollout, but with uneven maturity across business units, tools, and identities. That split creates governance debt because control assumptions change from one workflow to the next. Financial services should treat AI adoption as a distribution problem, not a single programme milestone.
Identity now sits on the critical path for AI governance: once employees move between SaaS copilots, desktop apps, and personal accounts, the organisation loses a stable map of who is doing what with which data. That makes IAM, account lifecycle control, and delegated access policy part of AI governance, not adjacent functions. Practitioners should expect identity boundaries to define where AI oversight succeeds or fails.
Split AI usage creates visibility asymmetry: the fastest adopters often accumulate the least predictable control surface because usage expands before inventory, policy enforcement, and telemetry mature. That pattern is structurally dangerous in regulated environments where the same data can move through sanctioned and unsanctioned paths in the same day. The governance lesson is to design for heterogeneous usage, not idealised standardisation.
Agentic AI widens the accountability gap if scope is not explicit: when an AI agent can act across multiple tools, a vague approval model no longer tells you who owns the outcome or how to reconstruct the path. This is where enterprise AI risk intersects with NHI governance most directly, because agent identity, privilege scope, and offboarding become security controls. Practitioners need accountable delegation before autonomous or semi-autonomous use scales.
Financial services should expect AI governance to converge with data security posture management: the real question is no longer only which model is approved, but where data can move, where identity is asserted, and where logging still follows the session. That makes cross-domain policy enforcement the deciding factor. The field is moving toward workflow-level governance, and teams should align controls accordingly.
From our research library:
- 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.
- Only 23% of IT leaders were very confident in their organisation's ability to manage security and governance for GenAI deployments, according to a 2025 Gartner survey of 360 IT leaders.
- Read next: AI Agent Identity Security Buyer's Guide
What this signals
Split adoption creates split control expectations: the organisations moving fastest will not be the ones with the simplest risk profile. As AI spreads across SaaS, endpoint, and unmanaged environments, security teams should expect more variance in policy enforcement and more difficulty proving who touched sensitive data. That is a governance design problem, not a tooling footnote.
AI governance now depends on identity boundaries: once user identity, device identity, and agent identity diverge, the control model has to follow the workflow rather than the application. Teams should prepare for a world where the same employee may use one governed account, one personal account, and one agent-mediated path in the same business process.
Agent ownership is the next control fault line: organisations that cannot name the owner, scope, and escalation path for agentic activity will not be able to govern outcomes at scale. That is where identity security and AI governance converge, and it is where programme owners should concentrate their next control reviews.
For practitioners
- Map every AI entry point Inventory GenAI SaaS, desktop AI apps, embedded AI features, and agent workflows so you can see where employees actually interact with models and data.
- Tighten control over personal and unmanaged accounts Block or review business use through personal or unmanaged accounts, especially where sensitive financial data may be copied into external AI tools.
- Classify agent-mediated activity separately Treat AI agents as a distinct governance class with named owners, explicit scope, and auditable delegation rather than folding them into generic application approvals.
- Extend monitoring to shadow AI paths Look for AI usage that bypasses central IT visibility and feed those paths into logging, DLP, and incident review workflows.
- Align identity and data controls Connect account lifecycle, access review, and data handling policy so governance follows the user, the agent, and the workflow together.
Key takeaways
- AI adoption in financial services is fragmenting across multiple tools and accounts faster than governance can keep up, which weakens visibility and control.
- The control problem is not AI itself but the mismatch between where data moves and where security teams can still observe and enforce policy.
- Teams need to govern AI entry points, delegated access, and unmanaged usage together or risk normalising exposure across everyday workflows.
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 and MITRE ATT&CK address 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 |
|---|---|---|
| NIST AI RMF | GOVERN — AI Governance and Accountability | The article centres on governance lag across enterprise AI adoption and risk management. |
| MAP — AI Risk Identification and Mapping | The report is fundamentally about where AI usage is happening and where exposure concentrates. | |
| Recommendation — Align AI oversight to GOVERN so ownership, policy, and accountability keep pace with adoption. Use MAP to inventory AI use cases, accounts, and data paths before you standardise controls. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Agentic workflows create identity and privilege issues when scope and delegation are unclear. |
| Recommendation — Apply ASI03 to constrain delegated scope and monitor privilege use in agent-driven workflows. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article highlights sprawling accounts and unmanaged access paths for AI usage. |
| Recommendation — Enforce PR.AA-05 so AI access stays tied to explicit entitlements and reviewed authorization. | ||
| MITRE ATT&CK | TA0006;TA0010 — Credential Access; Exfiltration | The risk pattern includes sensitive data movement across fragmented AI paths. |
| Recommendation — Map AI exposure scenarios to TA0006 and TA0010 to prioritise detection around credentialed data movement. | ||
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.
- Agentic AI: Autonomous AI systems capable of planning, deciding, and taking actions, including calling APIs, writing code, and orchestrating other agents, with minimal human oversight. Agentic AI introduces new NHI risks as agents must authenticate to external services.
- 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.
- Delegated Access: Delegated access is permission granted to one identity to act on behalf of another user, service, or system. In NHI environments, this usually appears in OAuth-connected apps and automation tooling. It is powerful, but it must be tightly scoped and reviewed because it can persist long after the original business need ends.
Deepen your knowledge
The NHI Foundation Level course covers NHI governance, agentic AI identity, and secrets management through the industry's only accredited NHI security programme. It helps security leaders connect identity controls to the broader AI and access decisions shaping their programmes.
Published by the NHIMG editorial team on August 2, 2026.
Updated on October 12, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org