The Act Rolls Out in Phases. Here's the Order.

One of the smartest things about the EU AI Act is that it didn't switch on all at once.

First came the bans. The prohibited practices took effect first, because the clearest harms needed the fastest response. Then came the duties for general-purpose AI models: documentation, transparency about training data, and safety evaluations for the most powerful systems.

And now the final phase is arriving. This summer, the full obligations for high-risk AI systems take effect: risk management, conformity assessments, human oversight, the whole homework list we've been talking about.

The order tells you how to plan. Handle what's already in force, prepare for what's next. Compliance is a sequence, not a scramble. Learn the order, and work it in the same order.

I explain frameworks; I don't give legal counsel.