The AI world isn’t standing still, and neither is the regulator anymore. On 2 August 2026, most of the EU AI Act will officially come into force. In this article you’ll read what will change, we share a handy step-by-step plan, and how we at aizy deal with the AI Act.

What is going to change?

What applies to you as an organisation depends on the risk class you fall into:

• Prohibited (in force since February 2025): Things like social scoring and subliminal manipulation are strictly prohibited. 

• High (from 2 December 2027): AI in healthcare, critical infrastructure, the judiciary, government services, and systems for recruitment & selection (e.g. AI that scans CVs or assesses applicants) or credit assessment. If you fall into this risk class, you must set up risk management systems, build in human oversight, keep extensive technical documentation and register the system in a central EU database before it may enter the market.

• Limited (from 2 August 2026): Most marketing and optimisation tools fall under ‘Limited risk’. This includes chatbots and AI image generation. You will soon be required to inform the end user that they are communicating with AI, or that content has been generated by AI. For AI-generated content, such as images and videos, it is not enough to simply add a text note; the law often also requires invisible, machine-readable metadata (watermarks) to be embedded in the file.

• Minimal: This includes spam filters and AI-powered spell checking. The AI Act adds no extra obligations or administrative burdens here.

The only obligation that always applies is AI literacy. This simply means that the employees within your organisation who work with AI must understand the basic principles, operation and risks of AI. The EU wants everyone who touches AI to do so consciously and competently.

Get started right away: your step-by-step plan for August

• Step 1: Make an AI inventory. Map out all AI tools within your organisation: from ChatGPT to chatbots.

• Step 2: Categorise by risk class. Determine which scale each tool falls under.

• Step 3: Train your staff. Make sure everyone is aware of the basic principles. 

• Step 4: Focus on transparency. Make sure your marketing materials from August clearly state when content has been generated by AI. Do you fall into a high risk class? Then you can already start preparing for the deadline of 2 December 2027. Use that time wisely, because building a watertight risk management system can easily take a year.

How does aizy deal with the AI Act?

With our AI solution, aizy falls under ‘Limited risk’. In that case, the obligation under the AI Act is mainly transparency. In your account, you can therefore always transparently see whether optimisations have been carried out automatically by AI, or whether a consultant has been at the controls.

Our customers' data is also never used to train public models from third parties. Everything that happens within aizy stays within the safe walls of our own infrastructure.

Our software combines real-time data from the Meta and Google APIs with our own optimised AI models. This allows us to deliver advice that is specifically tailored to your ad performance, without compromising the security of your business data.