AI as an insider threat is exposing gaps in risk models

At a glance

What this is: Gurucul's 2026 Insider Risk Report argues that AI has become an insider risk factor inside enterprise trust boundaries, with incidents increasingly spanning human users, machine identities and AI-driven activity.

Why it matters: This matters because insider governance is shifting from watching isolated events to controlling continuous identity behaviour across human IAM, NHI, and autonomous workflows.

By the numbers:

• 90% of organisations experienced at least one insider incident in the past 12 months, according to Gurucul.
• 94% of organisations say AI adoption is increasing their insider risk exposure, according to Gurucul.
• 54% of organisations acknowledge confirmed or suspected AI-related insider incidents, according to Gurucul.

Context

The article argues that insider risk is no longer limited to a discrete malicious or negligent user event. It now spans human users, machine identities, and AI-driven activity that can move at machine speed inside trusted enterprise workflows.

For identity teams, the governance problem is not just detection volume. It is that access, behaviour, and decision-making are now distributed across collaboration tools, cloud platforms, and shadow AI in ways older insider models do not represent well.

The article's central claim is that AI is being treated as an insider risk class in its own right, which is atypical for older insider risk programmes that were designed around people first.

Key questions

Q: What breaks when insider-risk programmes only monitor people and not AI agents?

A: Programmes miss the delegated layer where sensitive actions are increasingly executed. An AI agent may access data, chain tasks, or trigger workflows on behalf of a user, but if the agent is not tracked as an identity, its permissions and behaviour sit outside normal governance. That creates blind spots in accountability, review, and containment.

Q: Why do machine identities increase financial control risk?

A: Machine identities increase financial control risk because they can operate with standing privilege, scale across systems and keep acting after the original business need has changed. In finance workflows, that can allow silent supplier changes, record updates or payment actions without the friction that normally slows human abuse. The risk is hidden privilege plus high velocity.

Q: How do security teams know if insider risk monitoring is actually working?

A: Look for fewer isolated alerts and more explainable investigations that end in proportionate action. A working programme can show which signals were correlated, which cases were dismissed for legitimate context, and which interventions happened before data loss or excessive privilege use.

Q: How should security teams investigate insider risk when AI interactions are part of the communications trail?

A: Security teams should treat AI prompts, file attachments, responses, and downstream actions as part of the same investigative record as human messages. The goal is to reconstruct intent and context, not just detect a flagged event. Correlating AI activity with communications, data, and logs helps investigators decide whether behavior was routine, careless, or part of a broader pattern.

Technical breakdown

Why insider risk models break when AI behaves like an insider

Traditional insider risk models assume a human actor whose behaviour is slow enough to observe, classify, and investigate. AI changes that baseline because it can operate inside email, documents, workflows, and identities at machine speed while still looking like ordinary enterprise activity. That creates a governance mismatch: the control stack may still see events, but not the behavioural pattern that links those events into insider risk. When AI is granted delegated authority, the relevant question becomes not whether an action was allowed in isolation, but whether the actor can behave like an insider across multiple systems without leaving a clear human accountability trail.

Practical implication: monitor AI activity as an identity and behaviour problem, not only as an application telemetry problem.

How machine identities widen the insider threat surface

The report places machine identities alongside human users and AI-driven activity because insider risk now extends into accounts, tokens, workflows, and service access that can be reused or over-scoped. This is important because machine identities often lack the social cues and change patterns that human-focused monitoring relies on. Once those identities participate in workflows, they can amplify lateral movement, data access, and process abuse without looking like a traditional compromise. The governance failure is not simply that machine identities exist, but that they are often treated as operational plumbing rather than as part of the insider threat surface.

Practical implication: bring machine identity inventory, privilege review, and behavioural monitoring into insider risk operations.

What tool sprawl does to insider detection quality

The report says one-third of organisations run five or more insider risk tools, yet many still struggle with detection accuracy and data fragmentation. That points to a structural problem: separate tools may produce more alerts, but they do not necessarily create a coherent picture of identity behaviour across cloud, collaboration, and AI workflows. When context is split across systems, the programme can detect fragments of risk but miss the sequence that makes those fragments meaningful. In insider risk, poor correlation is not a tuning issue alone; it is a design issue that affects investigation speed and response confidence.

Practical implication: consolidate identity, behaviour, and access signals into a single investigation path before expanding alert volume.

Threat narrative

Attacker objective: The objective is to exploit trusted access paths from inside the organisation long enough to cause financial loss, data exposure, or process abuse without early containment.

1. Entry occurs through legitimate access inside email, documents, workflows, or identities, where AI and human activity blend into trusted enterprise use.
2. Escalation happens when delegated authority, machine identities, or shadow AI expand the scope of what the actor can see, touch, or automate.
3. Impact follows when risk is recognised too late to prevent material loss, especially where insider incidents are costly and containment lags behind detection.

Breaches seen in the wild

• Rabbit R1 hard-coded API keys 2024: Rabbit R1 source code leaked by an employee held working ElevenLabs, Azure, Yelp and Google Maps API keys; an email key was misused, Rabbit says.
• Ledger Connect Kit npm compromise 2023: Attackers phished a former Ledger employee with unrevoked npm access and published a wallet-draining Connect Kit, stealing about $600k on 14 Dec 2023.

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 is no longer just an insider risk enabler, it is becoming an insider risk subject. That matters because the programme assumption that insiders are either human or clearly malicious no longer fits the operating reality described here. When AI sits inside delegated authority, the identity boundary itself becomes part of the risk surface, and insider governance has to reflect that.

The real governance shift is from event review to behaviour governance. Legacy insider programmes were built to detect policy violations or suspicious acts after the fact. Continuous AI-driven activity makes that too slow, because the meaningful signal is the sequence of actions across systems, not the event in isolation.

Identity fragmentation is now a detection problem, not just an architecture inconvenience. The report shows that tool sprawl and data fragmentation are suppressing detection accuracy, which means the organisation may have more signals but less certainty. Practitioners should treat correlation quality as a core insider risk control, not a reporting enhancement.

Named concept: insider trust boundary collapse. The article describes a world in which humans, machine identities, shadow AI, and delegated AI authority all operate inside the same trust zone. That collapses the old idea that the insider boundary is primarily about employment status or user intent, and forces teams to govern access as a dynamic behaviour model.

AI agentic activity and machine identity governance are converging inside insider risk programmes. The report's strongest implication is that insider teams can no longer stay separate from NHI and autonomous governance conversations. If the programme cannot govern non-human access as part of insider risk, it will miss the fastest-moving part of the problem.

What this signals

Insider trust boundary collapse: programmes built around employee intent no longer cover AI-driven activity that executes inside trusted workflows. The practical implication is that insider monitoring has to follow delegated authority, not just user status or login source.

The strongest signal for practitioners is that behavioural analytics now has to sit alongside identity governance, not downstream from it. When machine identities, copilots, and shadow AI all participate in the same workflow, the investigation model must unify access, action, and context before containment decisions are made.

For practitioners

• Map AI and machine identities into insider risk scope Inventory copilots, generative AI tools, service accounts, tokens, and automated workflows as insider-risk actors, not just infrastructure components.
• Unify behavioural and access telemetry Correlate identity events, workflow activity, and content access in one investigation path so analysts can see sequences rather than isolated alerts.
• Re-tune insider use cases for continuous activity Shift detection logic from single policy violations to behavioural patterns that unfold across cloud platforms, collaboration tools, and AI-assisted work.
• Prioritise response paths for delegated authority Define containment steps for AI-driven and machine-driven activity before an investigation starts, including privilege suspension and workflow interruption.

Key takeaways

• AI-driven insider risk is not a new alert type, it is a change in what counts as an insider inside the enterprise.
• The report says 90% of organisations experienced insider incidents and 54% reported confirmed or suspected AI-related cases, which shows the problem is already operational.
• Teams need a unified view of human access, machine identities, and delegated AI behaviour if they want to contain insider risk before it spreads across workflows.

Key terms

• Insider Trust Boundary Collapse: A loss of the old boundary that separates trusted internal users from risky external actors. In modern environments, AI tools, machine identities, shadow IT, and delegated workflows can all operate inside the same trust zone, so the insider problem becomes behavioural rather than purely employment-based.
• Delegation authority: Delegation authority is the right for one identity to act on behalf of another within a defined scope. For agents, it must be explicit, time-bound, and auditable because the system may initiate actions independently once granted access.
• Behavioural Analytics: Behavioural analytics compares current activity against normal patterns to detect anomalies that may indicate abuse or compromise. In identity programmes, it is used to spot suspicious access behaviour that rule-based monitoring can miss, especially when attackers mimic legitimate workflows.
• Machine Identity: The digital identity of a machine, device, or workload, such as a server, container, or VM, used to authenticate it within a network. Sometimes used interchangeably with NHI, though NHI is the broader category.

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 identity security programme, it is worth exploring.