The EU AI Act is in force, and its obligations apply in stages. Many teams react with one of two reflexes: ignore it or freeze. Both are wrong. For most engineering teams, the AI Act is mainly a prompt to document properly what they should be doing anyway.
1. Most systems are not high-risk
The AI Act distinguishes by risk: prohibited practices, high-risk systems, systems with transparency obligations, and everything else. A coding assistant, an internal knowledge chatbot or a text summariser will usually not be high-risk. High-risk mainly concerns areas such as hiring decisions, creditworthiness, critical infrastructure or safety-relevant product components.
First step: build an inventory. Which AI systems do we develop or use - and for what?
2. Your role determines your obligations
Whoever develops an AI system and places it on the market under their own name is a provider and carries most of the obligations. Whoever only uses it is a deployer. A team that integrates a foundation model via API into its own product can become a provider itself.
3. Transparency is almost always required
When people interact with an AI system, they need to know. Generated content such as images, audio or video must be recognisable as such. Technically simple - if you plan for it from the start.
4. Documentation becomes part of the product
For high-risk systems, the AI Act requires technical documentation, risk management, logging and human oversight. The good news: most of that is engineering best practice. Teams that already record architecture decisions, training data, tests and monitoring are well ahead.
5. AI literacy affects everyone
Since February 2025, companies must ensure that employees working with AI have sufficient competence. It’s not rocket science, but it needs a plan - and evidence.
What to do now
- Inventory AI systems and assign a risk class
- Clarify roles: provider or deployer?
- Build transparency notices into UI and output
- Add documentation to your definition of done
- Keep an eye on the timeline - deadlines and guidance keep being refined
The AI Act doesn’t slow down teams that work well. It slows down those who don’t know what they’ve built.
This article is not legal advice.
