Currently on hold — not for sale. This product is under an internal accuracy review and has been withdrawn from promotion while that review is open. It cannot be bought at any amount while the hold is active; the Gumroad listing is set to $0 with an on-hold notice until it is unpublished there.

Guide

EU AI Act Compliance for SMEs: What You Need to Know in 2026

The EU AI Act's high-risk deadline moved to December 2027. Here's what SMEs should do with that runway — and what's already enforced.

Published September 2026 · 8 min read · Updated to reflect 2026 Digital Omnibus amendments
Disclosure: Galaxy Forge's related EU AI Act Compliance Toolkit is on commercial hold and is not for sale. This article is the information itself, free to use. Procedure information, not legal advice, and not a substitute for a qualified adviser — your obligations depend on the system you actually deploy. 0 sales recorded for the kit to date; no reviews.

Regulatory update: On August 2, 2026, only Article 50 transparency obligations (chatbot disclosure, deepfake labeling) took effect. The 2026 Digital Omnibus on AI deferred standalone high-risk system obligations from August 2, 2026 to December 2, 2027, and Annex I embedded systems from August 2, 2027 to August 2, 2028. This article reflects the current timeline.

What happened on August 2, 2026?

The EU AI Act entered its first enforcement phase. But not the phase most SMEs feared. The obligations that actually took effect are the narrower Article 50 transparency rules:

  • Chatbot disclosure — AI systems interacting with humans must clearly identify themselves as artificial
  • Deepfake labeling — AI-generated content that looks real must be labeled as such
  • Emotion recognition disclosure — systems reading emotions must inform the person being scanned

These are real obligations with real enforcement. But they're narrow. If your AI system doesn't fall into these categories, you have more time.

What got deferred to December 2027?

The bigger obligations — the ones that require real work — were pushed back 16 months:

  • Risk classification — determining whether your AI system is prohibited, high-risk, limited-risk, or minimal-risk
  • Technical documentation — Annex IV requirements for system design, testing, and monitoring
  • Conformity assessment — the formal process of demonstrating compliance for high-risk systems
  • Human oversight requirements — designing systems that allow meaningful human intervention
  • Data governance — quality and bias management for training data
  • Post-market monitoring — ongoing surveillance after deployment

These are the obligations that cost time and money. The question is: what do you do with 16 months of runway?

The SME problem

Most EU AI Act guides are written for enterprises. They assume you have a legal team, a compliance department, and a budget for consultants. SMEs have none of that.

The typical compliance consulting engagement costs €5,000–€50,000. For an SME, that's not a line item — it's a quarter of the compliance budget, if there even is one.

The gap isn't knowledge. It's practical implementation tools: templates, checklists, decision trees, and procedures that translate the regulation's 450+ pages into actionable steps.

What you should do now

1. Classify your AI systems

Before December 2027, you need to know what tier each of your AI systems falls into. Use the EU AI Act's risk classification framework:

  • Prohibited — social scoring, real-time biometric surveillance, emotion recognition in workplaces/schools
  • High-risk — AI in critical infrastructure, education, employment, law enforcement, migration
  • Limited-risk — chatbots, deepfakes, emotion recognition
  • Minimal-risk — spam filters, AI-enabled video games

Classification drives everything else. Get this wrong and you either over-invest in compliance you don't need, or under-invest and face penalties.

2. Document what you have

Technical documentation isn't optional. Annex IV specifies what you need: system description, design specifications, development methodology, validation data, risk management, and monitoring procedures.

Start now. Documentation written after the deadline is documentation that's rushed, incomplete, and expensive.

3. Build human oversight

High-risk AI systems must allow meaningful human intervention. This means more than a kill switch — it means designing workflows where humans can understand, monitor, and override the system's decisions.

4. Prepare for conformity assessment

High-risk systems need a conformity assessment before deployment. For most SMEs, this means internal assessment with documentation — not a notified body engagement. But you need the documentation first.

The deadline is real, but so is the runway

December 2, 2027 is not a suggestion. Penalties for non-compliance reach €35 million or 7% of global annual turnover, whichever is higher. For an SME, the lower absolute number is still significant.

But 16 months is real time. Companies that start now — with practical tools, not expensive consultants — will be audit-ready well before the deadline.

What a practical toolkit looks like

You don't need a 450-page legal analysis. You need:

  • A risk classification decision tree for your specific AI systems
  • Documentation templates mapped to Annex IV
  • A conformity assessment procedure you can follow
  • Human oversight design requirements
  • An audit-readiness checklist
  • A 90-day implementation roadmap

These tools don't replace legal advice. They make your legal spend targeted instead of exploratory. A consultant reviews what you've built; they don't build it from scratch.

EU AI Act Compliance Toolkit - on hold

The related toolkit is on commercial hold and is not for sale. This article remains free to use.

On hold - not for sale

It cannot be bought at any amount while the accuracy review is open. If you were ever charged for this product, email galaxyaek7@gmail.com and we will refund you in full.

Disclaimer: This article and toolkit are informational, not legal advice. EU AI Act compliance depends on your specific AI systems, use cases, and member state implementation. Consult a qualified legal professional for your specific situation.