Cyber Horizon
Back to Blog
EU AI ActAI GovernanceEU

EU AI Act: A Practical Compliance Guide for 2026

17 June 2026·9 min read·Cyber Horizon Team

The EU AI Act is the world’s first comprehensive law governing artificial intelligence. Like GDPR before it, its reach extends well beyond Europe: if your AI system is used by people in the EU, you are in scope — wherever you are based.

Here is what the Act actually requires, how it classifies AI by risk, and the practical steps to get ahead of it.

A risk-based law, not a blanket ban

The Act regulates AI in proportion to the harm a system can cause. Every system you build, deploy, or distribute falls into one of four tiers — and your obligations follow from there.

Risk tierExamplesObligation
UnacceptableSocial scoring, manipulative or exploitative AIProhibited
HighHiring, credit, biometric ID, critical infrastructureStrict requirements
LimitedChatbots, deepfakes, emotion recognitionTransparency duties
MinimalSpam filters, recommendation engines, most softwareNo new obligations

Who is in scope?

The Act distinguishes between providers (who develop or place an AI system on the market) and deployers (who use one in a professional context). Both carry obligations. Most organisations are deployers — and many underestimate the duties that come with simply using a high-risk system bought from a vendor.

What high-risk systems must do

Risk management system

A continuous, documented process to identify and mitigate risks across the AI lifecycle.

Data governance

Training, validation and testing data must be relevant, representative and checked for bias.

Technical documentation

Evidence the system meets the requirements — maintained and available to regulators.

Record-keeping & logging

Automatic logs to ensure traceability of the system’s functioning.

Human oversight

Designed so a person can understand, intervene in, and override the system.

Accuracy, robustness & cybersecurity

Appropriate levels for the intended purpose, resilient to error and attack.

Deadlines that matter

The Act applies in phases. Prohibited-use rules and AI-literacy duties landed first; obligations for general-purpose AI models followed; and the full high-risk regime phases in through 2026 and 2027. The penalties are GDPR-scale — up to €35M or 7% of global turnover for prohibited uses — so the planning horizon is now, not when enforcement bites.

How to prepare

  • Inventory every AI system you build or use, and classify each by risk tier.
  • Map your high-risk and limited-risk systems to the relevant obligations.
  • Stand up an AI governance function — ideally aligned to ISO 42001.
  • Document data sources, testing, and human-oversight controls as you go.
  • Add transparency notices wherever users interact with AI.

The bottom line

The EU AI Act turns responsible-AI principles into legal obligations with real teeth. Treat it like you treated GDPR: inventory what you have, classify by risk, and build the governance and evidence trail now — while the deadlines are still ahead of you.

Govern AI alongside the rest of your programme

Cyber Horizon maps the EU AI Act and ISO 42001 to your controls, tracks evidence, and keeps your AI governance audit-ready — in the same platform as the rest of your GRC.

Book a Demo