Your guide to the EU AI Act

12 July 2024: The AI Act hits the books

Mark your calendars—this is when the AI Act was officially published in the EU’s Official Journal. Think of it as the formal “heads-up” to everyone involved, setting the stage for what’s coming next. The clock starts ticking from here!

1 August 2024: The AI Act becomes an official law

This was the day the AI Act went from paper to reality. While you won’t see any immediate changes, from this day forward AI providers and users should start getting their ducks in a row. The Act is now officially on the books, so it’s time to prepare for the new regulations coming your way.

2 February 2025: Say goodbye to prohibited systems

Starting as of this date, certain AI systems are getting the boot. We're talking about tech that the EU has labelled as too risky; like government social scoring, some types of deepfakes, and real-time biometric surveillance in public. From now on, using or creating these kinds of AI is a no-go across the EU.

2 August 2025: General-purpose AI models and systems get their regulations

Six months later, the spotlight turns to general-purpose AI models. This includes the AI behind systems such as chatbots, image recognition, and much more. Providers of these systems will need to start playing by the rules, making sure their tech is safe, transparent, and reliable. This means that human oversight and solid accuracy standards are an absolute must.

2 August 2026: High-risk AI systems and sandboxes come into play

Fast forward to August 2026, and we’re moving into high-risk territory. This includes AI systems used in critical areas like healthcare and law enforcement. The EU is tightening the screws with tougher regulations; think rigorous testing, risk management, and constant monitoring. Additionally, this date also introduces “regulatory sandboxes”. These are special zones where companies can test their AI in a controlled environment, all under the watchful eye of regulators.

2 August 2027: More high-risk AI systems face the music

By August 2027, even more high-risk AI systems come under the microscope. This final phase targets systems that are crucial to public services and fundamental rights, such as AI used in emergency responses or critical infrastructure. The regulations for these are all about ensuring the systems are super safe, well-documented, and transparent.

Wrapping it up

The EU AI Act is rolling out gradually, giving everyone time to adjust to the new regulations. AS of now until 2027, businesses and organisations need to stay on top of these deadlines and make sure their AI tech is up to par. The future of AI in Europe is taking shape, make sure you are ready for it!