Skip to main content
    Guide
    12 min read

    AI Inventory for EU AI Act Compliance

    Your AI inventory is the foundation of compliance. Learn what to capture, how to maintain it, and the mistakes that trip up most organizations.

    Why Your AI Inventory Matters

    Single Source of Truth

    You can't classify, control, or evidence compliance for systems you don't know about.

    Risk Visibility

    See which systems are high-risk, which need transparency notices, and where gaps exist.

    Audit Evidence

    When regulators or customers ask, your inventory is the first thing they'll want to see.

    Essential Fields to Capture

    A compliant AI inventory needs these data points for each system.

    System Name & ID
    Required
    Unique identifier for tracking across your organization
    Business Owner
    Required
    The person accountable for compliance decisions
    Vendor/Provider
    Required
    Who built it—internal, SaaS, API, or open-source
    Purpose & Use Case
    Required
    What decision or output does it influence?
    Affected Groups
    Required
    Customers, employees, candidates, students, etc.
    Deployment Region
    Required
    EU countries where the system is used
    Risk Classification
    Required
    Minimal, limited, high-risk, or prohibited
    Human Oversight
    HITL, HOTL, or HOOTL model
    Data Types Processed
    Personal data, special categories, minors
    Logging & Retention
    Where logs are stored and for how long

    How to Build Your Inventory

    1

    Identify all AI systems

    Audit your software stack for any system that infers outputs from inputs—chatbots, ML models, recommendation engines, automated decision-making tools.

    2

    Gather system information

    For each system, document the vendor, purpose, affected groups, deployment regions, and data types processed.

    3

    Assign ownership

    Designate a business owner for each system who is accountable for compliance decisions.

    4

    Classify risk levels

    Use the Annex III criteria to determine if each system is minimal, limited, or high-risk.

    5

    Document controls

    Record human oversight models, logging capabilities, and operational procedures.

    6

    Establish review cadence

    Set triggers for quarterly reviews, model updates, and vendor changes to keep the inventory current.

    Common Mistakes to Avoid

    Only listing 'obvious' AI

    Chatbots, recommendation engines, automated screening tools are all AI. Audit your entire software stack.

    Using spreadsheets without audit trails

    Excel doesn't capture who changed what and when—critical for compliance evidence.

    Set-and-forget mentality

    AI systems evolve. Your inventory needs regular review triggers for model updates, new use cases, and vendor changes.

    No ownership assignment

    Every system needs a named business owner. 'IT' or 'The team' isn't good enough.

    Related Resources

    Inventory Template

    Downloadable spreadsheet with all required fields.

    Download Template

    Definition Checker

    Not sure if it's an AI system? Use our checker.

    Check Now

    Inventory Software

    Automated inventory with audit trails and exports.

    Learn More

    Frequently Asked Questions

    What is an AI system inventory under the EU AI Act?

    An AI system inventory is a documented register of all AI systems your organization uses or deploys. It captures essential information like system purpose, risk classification, ownership, and compliance status—forming the foundation of EU AI Act compliance.

    Why is an AI inventory the first step in compliance?

    You can't classify, control, or evidence compliance for systems you don't know about. The inventory is your single source of truth that enables all other compliance activities—from prohibited practice screening to high-risk deployer obligations.

    What fields should an AI inventory capture?

    At minimum: system name, owner, vendor, purpose, affected groups, deployment regions, risk classification, and human oversight model. Additional fields like data types, logging, and transparency requirements help with Article 26 and 50 obligations.

    How often should we update the AI inventory?

    Review quarterly at minimum. Additionally, update immediately when there are model changes, new use cases, vendor updates, or material changes that could affect risk classification.

    Can we use a spreadsheet for our AI inventory?

    While spreadsheets work initially, they lack audit trails, automated classification, and evidence linking. Purpose-built tools like Klarvo provide these features essential for demonstrating compliance to auditors.

    Ready to Build Your AI Inventory?

    Klarvo automates inventory management with guided intake, classification, and audit-ready exports.

    No credit card
    14-day trial
    Cancel anytime