Index / Methodology

How the index is scored

The index turns a short questionnaire into one exposure score across five weighted signals. The scoring is transparent, and the full question set publishes with the assessment.

The five signals

Your answers are scored across five signals. Each carries a weight set by how much that control, present or absent, moves real exposure. A visibility gap raises the score more than an audit gap, which is why the two are not weighted the same.

AI visibilityweight 25
Data movementweight 25
Identity and ownershipweight 20
Evidence and auditweight 15
Remediation reachweight 15
  • AI visibility. A live inventory of every AI tool, agent, and connected account in use, corporate and personal, or a set of blind spots you are guessing at.
  • Data movement. Whether sensitive content is caught the moment it moves toward an AI tool, before it lands in a model.
  • Identity and ownership. Whether every AI action ties back to the person driving it, with their role, status, and an accountable owner.
  • Evidence and audit. Whether you keep a per-prompt record that would hold up in front of an auditor.
  • Remediation reach. Whether you can act through the stack you already run, revoke, quarantine, uninstall, without standing up a new agent.

How the score is calculated

Each answer is worth exposure points from 0 to 100, where 0 means the control is fully in place and 100 means it is absent. A signal subscore is the average of its questions. The index is the weighted sum of the five subscores, so it also runs 0 to 100, and higher means more exposed.

index = visibility×0.25 + data movement×0.25 + identity×0.20 + evidence×0.15 + remediation×0.15
Higher score means more exposed. Lower is better.

This is a self-reported, directional read, not an audit. It is built to be consistent, so the number you record this quarter is comparable to the one you record next.

The five exposure bands

The score places an organization in one of five bands. The bands run from controls holding across the board to little visibility or control at all.

Contained0 to 20
Guarded21 to 40
Elevated41 to 60
Exposed61 to 80
Critical81 to 100
  • Contained (0 to 20). Controls hold across the five signals. Keep them current as AI use grows.
  • Guarded (21 to 40). The basics are in place. A few gaps are worth closing before they widen.
  • Elevated (41 to 60). Real exposure sits in more than one signal. This is where most teams are today.
  • Exposed (61 to 80). Sensitive data can move through AI with little to stop it. Work the weakest signals first.
  • Critical (81 to 100). There is little visibility or control over AI-era data movement. Start with the fundamentals.

Where the numbers come from

Every statistic on this site is attributed to a primary source, and we cite only major research organizations and established players in the field. We do not cite security startups or vendors marketing a competing product. The figures behind the index today are drawn from:

  • Microsoft and LinkedIn, 2024 Work Trend Index
  • Verizon, Data Breach Investigations Report (2025 and 2026)
  • Gartner research and forecasts on AI and security, 2025
  • IBM, Cost of a Data Breach Report 2025
  • Ponemon Institute, 2026 Cost of Insider Risks Global Report

Where a number appears on this site, its source appears next to it. We never invent a figure or round past what the source says.

Who is behind the index

The AI Insider Risk Index is built and maintained by Anzenna, an AI security company. We say that plainly because a benchmark is only useful if you can trust it. The scoring never asks about or rewards any vendor's product, Anzenna's included. More on how we handle this is on the about page.

15 questions · ~10 min · no registration

Score your exposure

Fifteen questions, about ten minutes, no registration to see your result.

Run the assessment