A meaningful institutional test
The EU AI Act entered into force in 2024, and important enforcement powers became applicable in August 2026. Other requirements, including parts of the high-risk regime, follow different timelines. It is inaccurate to say every AI Act obligation applies to every system today.
Rules need evidence
A standard matters when regulators can request technical information, examine real-world harms and enforce consequences. Businesses need legal clarity; people need safety, transparency and a route to contest important decisions.
The tension is real
Badly designed regulation can freeze smaller entrants out of markets or create paperwork without improving safety. Weak regulation can shift hidden risks to the public. The answer is proportionate obligations, independent tests and honest evaluation of what works.
The next milestone
Implementation should be judged by outcomes: dangerous failures detected, rights exercised, safer procurement and increased public trust. Citizens can ask lawmakers to publish enforcement capacity, incident data and the results of independent evaluations.
Judge governance by whether it reduces harm and improves recourse — not by the number of pages in a rulebook.