The Council gives final approval to simplify and streamline the EU's AI rules
The Council of the EU gave its final green light to a package simplifying and streamlining the AI rulebook — arriving in the same year enforcement powers switched on. Brussels is now tightening and loosening the same regime simultaneously, which tells you something about how the first year of implementation went.
A legislature that has just armed a regulator does not usually spend the same session making the rules easier to comply with. That the Council has done both is the most honest available signal about the gap between the Act as drafted and the Act as implementable.
The simplification package is procedural rather than substantive — the prohibited practices are not coming back, and the high-risk annex still binds. What moves is the machinery around it: documentation burden, sequencing, and the administrative load on smaller deployers who were never the intended target of a regime designed for frontier systems and safety-critical deployments.
The tension is structural and will not resolve. The Act has to be heavy enough to bite on the handful of actors capable of systemic harm and light enough not to strangle the thousands of firms that merely bought a model from one of them. Those two requirements pull in opposite directions, and every jurisdiction writing AI law is discovering it at roughly the same point in the implementation curve.
For firms inside scope, the operational reading is unchanged: the high-risk obligations are in force, and simplification is about how you evidence compliance, not whether you owe it.
Council of the EU — Artificial intelligence: Council gives final green light to simplify and streamline rules → · European Commission — AI Act | Shaping Europe's digital future →