TROIE
The EU AI Act becomes enforceable on August 2, 2026Learn more →The EU AI Act becomes enforceable on August 2, 2026Learn more →The EU AI Act becomes enforceable on August 2, 2026Learn more →The EU AI Act becomes enforceable on August 2, 2026Learn more →The EU AI Act becomes enforceable on August 2, 2026Learn more →The EU AI Act becomes enforceable on August 2, 2026Learn more →
← The Journal
AI at work7 min21 June 2026

EU AI Act: what your company needs to do in 2026

Your teams are already using AI. The question is no longer “should we adopt it” but “are we compliant”. Since 2025, Europe has had a framework in place, the EU AI Act, and some of its obligations become binding in 2026. Here is what that means for your company in practice, without the legal jargon.

The EU AI Act in one minute

The EU AI Act is the European regulation on artificial intelligence. Its core principle: the riskier the use of AI, the stronger the obligations. Most companies do not build AI, they use it, so what mainly concerns them comes down to two things: transparency and AI literacy, meaning that your teams have a minimum level of competence to handle these tools.

The obligation that really applies to you: training your teams

This is the point many executives miss. The regulation requires that people who use AI systems at work have a sufficient level of competence to do so in an informed way. In other words: letting your employees improvise with ChatGPT, with no framework and no training, is no longer a neutral option.

And the deadline is closing in: national penalties for non-compliance become applicable in 2026. The good news is that getting compliant on this front is entirely manageable: it is mostly a matter of setting a framework and training.

The real risk in the meantime: shadow AI

Today, in most companies, AI comes in through the back door: one employee pastes a confidential document into a consumer tool, another generates a contract without having it reviewed, a third circulates a flawed analysis because the AI “hallucinated”. This is what we call shadow AI: usage that is massive, yet invisible and unmanaged.

A team improvising with AI is not productivity, it is risk (data, compliance, quality).

The 4 steps to get compliant (and more effective)

  1. Take stock. Which tools are already in use, by whom, and for what? You can only manage what you know about.
  2. Set simple rules. What can and cannot be pasted, what must be reviewed, which tools are approved. A one-page policy is enough to start.
  3. Train your teams. Not a two-day theoretical seminar, but practical upskilling: prompting well, verifying outputs, protecting data, keeping a human in the loop.
  4. Document it. Keep a record of what you have done (policy, training completed). That is what proves your good faith.

Why this is an opportunity, not just a constraint

The numbers speak for themselves: AI use at work doubled in a year, yet more than half of employees have never been trained. Companies that train their people gain a real productivity edge, and AI skills can significantly raise the value of a role. Getting compliant also means getting better. It is the same underlying shift described in Manager, not replaced.

Where to start

The simplest move is an honest assessment of where you stand. At TROIE, we help teams frame and use AI, on site or remotely, with real teaching and zero fluff: see TROIE for professionals, or book a 30-minute call to review your compliance and your needs.

This article is informational and educational; it does not constitute legal advice. For an analysis of your specific situation, consult a specialised adviser.