Shadow AI governance gaps expose sensitive data at scale in Australia

At a glance

What this is: This is a Josys survey on shadow AI in Australian workplaces, showing that sensitive data is being uploaded to AI tools faster than organisations can govern it.

Why it matters: It matters because unmanaged AI usage creates a governance blind spot across human identity, access policy, and data handling, which security and IAM teams must close before compliance and exposure widen.

By the numbers:

• 36% of employees upload sensitive company information to AI tools.
• 70% of organisations have moderate to no visibility into what AI tools are being used.
• 63% of professionals lack confidence in using AI securely.

👉 Read Josys's report on shadow AI data exposure in Australia

Context

Shadow AI is the use of unauthorised AI platforms by employees outside formal security oversight. In this report, the core problem is visibility and control failure, not simply higher AI adoption. When staff can move strategy, financial, customer, and legal data into external tools without policy enforcement, identity governance breaks at the point of use.

For IAM and security teams, the issue sits across human access, acceptable use, and data handling controls. The article shows that organisations are trying to manage AI risk with manual review and incomplete tooling, even as employee behaviour moves faster than approval and monitoring processes. That gap affects compliance, audit readiness, and the trust boundary around sensitive information.

Key questions

Q: How should security teams govern employee use of external AI tools?

A: Security teams should treat external AI usage as a governance and data-control problem, not just an awareness issue. The practical approach is to inventory sanctioned and unsanctioned tools, classify the data they may receive, and enforce policy at the point of use with DLP, browser controls, and explicit acceptable-use rules.

Q: Why does shadow AI create risk even when users have valid corporate access?

A: Valid corporate access does not control where a user sends data after login. Shadow AI creates risk because the employee can remain authenticated while moving sensitive content into a third-party platform outside approved governance. That gap exposes confidential material even when authentication and role assignment are correct.

Q: What do organisations get wrong about AI governance policy?

A: Many organisations confuse written policy with effective control. A policy document does not stop data from being pasted into an external AI service, and it does not produce visibility into tool usage. Effective governance requires enforcement, monitoring, and a current inventory of approved AI services.

Q: Who should own shadow AI risk in an organisation?

A: Shadow AI risk should be owned jointly by IAM, security operations, privacy, and compliance teams because the issue spans identity, data handling, and regulatory exposure. If ownership sits in only one function, the organisation usually gets either weak enforcement or weak accountability, but not both.

Technical breakdown

Shadow AI as an identity governance problem

Shadow AI becomes an identity issue when users can authenticate to unsanctioned AI services with corporate accounts, personal logins, or unmanaged browser sessions and then transfer sensitive data outside approved control paths. The risk is not limited to the tool itself. It is the combination of identity, device, data classification, and policy enforcement failure. Without visibility into which tools are used and what data is entered, governance cannot distinguish benign experimentation from high-risk disclosure.

Practical implication: map sanctioned and unsanctioned AI use to identity, device, and data policies before the behaviour becomes normalised.

Why manual AI governance does not scale

Manual policy review cannot keep up with high-frequency AI usage across business units. Once employees can choose tools on demand, security teams need controls that evaluate risk at the point of access and at the point of data entry. This is closer to runtime policy enforcement than classic annual review. The report’s findings show that governance maturity is lagging adoption, which means approvals alone cannot contain the exposure.

Practical implication: move from document-based policy to enforceable controls tied to role, data sensitivity, and approved AI services.

Role-based access does not solve data leakage alone

Role-based access control can limit who may use approved systems, but it does not stop a user from pasting restricted information into a third-party AI platform. That gap matters because the control failure is at the boundary between user intent and data movement. In other words, access to the source system is not the same as control over the destination. Shadow AI exposes that mismatch directly.

Practical implication: pair access governance with data handling policy, DLP signals, and explicit AI usage rules.

NHI Mgmt Group analysis

Shadow AI is a governance failure before it is a technology problem. The Josys findings show that the primary breakdown is not user curiosity, but the absence of controlled visibility and enforceable policy around where sensitive information can go. When 70% of organisations lack clear sight of the tools in use, the programme cannot govern what it cannot enumerate. Practitioners should treat shadow AI as an access and data-control boundary issue, not a training-only issue.

Human identity controls do not cover prompt-based data leakage. Traditional user authentication confirms who entered the environment, but it does not govern what the user discloses after login. That distinction matters because employees are now able to move strategy, financial, customer, and legal data into external AI platforms in a single interaction. The implication is that identity programmes must extend beyond login assurance into usage policy and data egress control.

Policy without enforcement creates compliance theatre. The report’s mix of manual review, low preparedness, and weak enforcement tools shows a familiar failure mode: organisations can describe the policy but cannot consistently apply it. This is a classic gap between declared governance and operational governance. The practical conclusion is that AI governance must be measurable at the control point, not only documented in policy language.

Shadow AI accelerates the convergence of IAM, data security, and user behaviour analytics. AI usage is no longer a niche privacy concern because it touches access review, acceptable use, and sensitive data classification at the same time. That makes it a cross-domain governance problem that no single control family can solve in isolation. IAM teams, data security teams, and compliance leads need a shared operating model for sanctioned AI use.

Visibility is now the first control, not the last report. Organisations cannot meaningfully enforce role-based rules or sensitivity-based policy if they do not know which AI tools employees are using. The report’s central signal is that visibility lags behaviour, which means the governance baseline is already behind the workforce. Practitioners should treat discovery and inventory of AI use as a prerequisite for any effective control programme.

From our research:

• The average estimated time to remediate a leaked secret is 27 days, despite 75% of organisations expressing strong confidence in their secrets management capabilities, according to The State of Secrets in AppSec.
• Only 44% of developers are reported to follow security best practices for secrets management, according to The State of Secrets in AppSec.
• For teams building a broader control baseline, the Guide to the Secret Sprawl Challenge helps connect secret exposure with the governance gap shadow AI exposes.

What this signals

Shadow AI will force identity teams to treat approved tool lists as living controls. Once employees can route sensitive data into unauthorised services, annual policy review becomes too slow to matter. The operational signal to watch is whether discovery, classification, and blocking are happening continuously, not whether a policy exists on paper.

The governance gap also cuts across data security and access management. With 43% of security professionals concerned about AI systems learning and reproducing sensitive information patterns from codebases, organisations need to align AI usage controls with secrets handling and data loss prevention, not manage them as separate programmes.

A useful benchmark is whether your organisation can tie AI service approval to role, data class, and business function without manual exception handling. If it cannot, shadow AI is already operating as an unmanaged access channel rather than a fringe behaviour.

For practitioners

• Audit unsanctioned AI usage across the organisation Discover which AI tools are being used by department, identity type, and device class. Prioritise business units that handle strategy, financial, customer, or legal data, then document which tools are approved, tolerated, or prohibited.
• Enforce data sensitivity rules at the point of use Tie policy enforcement to data classification so that sensitive content is blocked or warned on before it reaches external AI services. Combine DLP, browser controls, and AI-specific acceptable use rules instead of relying on annual policy review.
• Replace manual review with measurable AI governance Track approvals, exceptions, blocked events, and policy violations as operational metrics. Use those signals to show whether governance is actually reducing exposure rather than simply producing documentation for audit.
• Align IAM and compliance teams on sanctioned AI services Maintain a current list of approved AI platforms, the identities allowed to use them, and the categories of data each service may receive. Review that list alongside access governance and privacy obligations so the control model stays current.

Key takeaways

• Shadow AI is exposing a structural governance gap because employees can move sensitive data into external AI tools faster than organisations can see or control the behaviour.
• The evidence points to a readiness problem as much as a usage problem, with broad adoption, low confidence in secure use, and limited visibility all appearing together.
• Practitioners should move from policy statements to enforceable controls that combine discovery, data classification, and sanctioned AI service management.

Key terms

• Shadow AI: Shadow AI is the use of AI tools, platforms, or services outside formal security approval and governance. In practice, it creates a blind spot because users can move sensitive data into systems that the organisation does not inventory, monitor, or control.
• AI Governance: AI governance is the set of policies, controls, and accountability mechanisms that define how AI tools may be used in an organisation. It becomes effective only when policy is backed by visibility, enforcement, reporting, and ownership across security, privacy, and IT.
• Data Classification: Data classification is the practice of identifying information by sensitivity so controls can match the risk. In shadow AI scenarios, classification determines whether content may be pasted into external tools, blocked entirely, or allowed only under approved conditions.

Deepen your knowledge

NHI governance, machine identity security, and secrets management are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity security strategy or governance in your organisation, it is worth exploring.

This post draws on content published by Josys: New Report Reveals That Over 1/3 of Australian Professionals Expose Sensitive Company Data to AI Platforms. Read the original.