August 9, 2026

Newsletter

The European Post

Policy • Diplomacy • Strategic Insight

LATEST
New EU Gas Market Rules Take Effect With Consumer SafeguardsDeforestation Regulation Holds Its December Deadline Despite CutsDigital Omnibus Delays Europe’s High-Risk AI Deadline to 2027Europe’s Schools Confront a Widening Basic Skills GapRule of Law Report Finds Progress but Trust Still LagsIRIS2 Gains 66 Satellites as Brussels Signs the Rollout DealNature Restoration Plans Come Due as Member States Race a DeadlineECB Rates Hold at 2.25 Percent as Frankfurt Waits on the DataEU Budget Talks Turn to Cohesion as Ireland Takes OverWildfire Response Stretches Europe’s Emergency Fleet ThinCan Europe’s Quantum Plan Match Its Big Ambitions?Armenia Edges West as Moscow’s Grip Slips AwayEurope Inherits the World’s Aid Bill It Cannot Fully PayAir Passenger Rights Get Their First Overhaul in Two DecadesErasmus Opens a 5 Billion Euro Round for Skills and MobilityCeuta Crisis Pushes the EU Toward Faster Migrant ReturnsBulky Parcels Face a Packaging Crackdown From August 12Can Europe Build Its Drone Wall Before 2030 Arrives?Indonesia’s Trade Deal With Europe Nears Its Final SignatureEurope Leans on Beijing to Keep Climate Talks Alive Before AntalyaNew EU Gas Market Rules Take Effect With Consumer SafeguardsDeforestation Regulation Holds Its December Deadline Despite CutsDigital Omnibus Delays Europe’s High-Risk AI Deadline to 2027Europe’s Schools Confront a Widening Basic Skills GapRule of Law Report Finds Progress but Trust Still LagsIRIS2 Gains 66 Satellites as Brussels Signs the Rollout DealNature Restoration Plans Come Due as Member States Race a DeadlineECB Rates Hold at 2.25 Percent as Frankfurt Waits on the DataEU Budget Talks Turn to Cohesion as Ireland Takes OverWildfire Response Stretches Europe’s Emergency Fleet ThinCan Europe’s Quantum Plan Match Its Big Ambitions?Armenia Edges West as Moscow’s Grip Slips AwayEurope Inherits the World’s Aid Bill It Cannot Fully PayAir Passenger Rights Get Their First Overhaul in Two DecadesErasmus Opens a 5 Billion Euro Round for Skills and MobilityCeuta Crisis Pushes the EU Toward Faster Migrant ReturnsBulky Parcels Face a Packaging Crackdown From August 12Can Europe Build Its Drone Wall Before 2030 Arrives?Indonesia’s Trade Deal With Europe Nears Its Final SignatureEurope Leans on Beijing to Keep Climate Talks Alive Before Antalya
August 9, 2026

Newsletter

The European Post

Policy • Diplomacy • Strategic Insight

LATEST
New EU Gas Market Rules Take Effect With Consumer SafeguardsDeforestation Regulation Holds Its December Deadline Despite CutsDigital Omnibus Delays Europe’s High-Risk AI Deadline to 2027Europe’s Schools Confront a Widening Basic Skills GapRule of Law Report Finds Progress but Trust Still LagsIRIS2 Gains 66 Satellites as Brussels Signs the Rollout DealNature Restoration Plans Come Due as Member States Race a DeadlineECB Rates Hold at 2.25 Percent as Frankfurt Waits on the DataEU Budget Talks Turn to Cohesion as Ireland Takes OverWildfire Response Stretches Europe’s Emergency Fleet ThinCan Europe’s Quantum Plan Match Its Big Ambitions?Armenia Edges West as Moscow’s Grip Slips AwayEurope Inherits the World’s Aid Bill It Cannot Fully PayAir Passenger Rights Get Their First Overhaul in Two DecadesErasmus Opens a 5 Billion Euro Round for Skills and MobilityCeuta Crisis Pushes the EU Toward Faster Migrant ReturnsBulky Parcels Face a Packaging Crackdown From August 12Can Europe Build Its Drone Wall Before 2030 Arrives?Indonesia’s Trade Deal With Europe Nears Its Final SignatureEurope Leans on Beijing to Keep Climate Talks Alive Before AntalyaNew EU Gas Market Rules Take Effect With Consumer SafeguardsDeforestation Regulation Holds Its December Deadline Despite CutsDigital Omnibus Delays Europe’s High-Risk AI Deadline to 2027Europe’s Schools Confront a Widening Basic Skills GapRule of Law Report Finds Progress but Trust Still LagsIRIS2 Gains 66 Satellites as Brussels Signs the Rollout DealNature Restoration Plans Come Due as Member States Race a DeadlineECB Rates Hold at 2.25 Percent as Frankfurt Waits on the DataEU Budget Talks Turn to Cohesion as Ireland Takes OverWildfire Response Stretches Europe’s Emergency Fleet ThinCan Europe’s Quantum Plan Match Its Big Ambitions?Armenia Edges West as Moscow’s Grip Slips AwayEurope Inherits the World’s Aid Bill It Cannot Fully PayAir Passenger Rights Get Their First Overhaul in Two DecadesErasmus Opens a 5 Billion Euro Round for Skills and MobilityCeuta Crisis Pushes the EU Toward Faster Migrant ReturnsBulky Parcels Face a Packaging Crackdown From August 12Can Europe Build Its Drone Wall Before 2030 Arrives?Indonesia’s Trade Deal With Europe Nears Its Final SignatureEurope Leans on Beijing to Keep Climate Talks Alive Before Antalya

How The AI Office Plans To Police Europe’s Most Powerful Models

Fyodor Dostoevsky
Policy Tracker, Digital and Technology

Brussels: Almost a year after the Artificial Intelligence Act entered into force, the law has stopped being an abstraction and started being a deadline. The first prohibitions, on practices such as social scoring and untargeted facial-recognition scraping, took effect in February 2025. The rules for general-purpose models, the engines behind today’s chatbots and image generators, followed that summer. The next, larger wave, covering most high-risk systems used in hiring, credit, education and critical infrastructure, lands in 2026. Each step pulls the regulation closer to the messy reality of enforcement.

At the centre of that effort sits the AI Office, a unit housed inside the Commission’s digital department and tasked with supervising the most capable systems directly. Its remit is unusual. National authorities will police most uses of AI within their borders, but the Office alone oversees the handful of foundation models judged powerful enough to pose systemic risk, those trained with computing power above a threshold the law fixes in raw arithmetic. For the first time, a European regulator is being asked to look inside the models themselves rather than only at how companies deploy them.

The instrument it leans on is the General-Purpose AI Code of Practice, a voluntary rulebook drafted with industry, academics and civil society. Signing up is meant to be the path of least resistance: firms that follow the code gain a presumption of compliance and a degree of legal certainty, while those that refuse must demonstrate conformity by other, harder means. The approach is pragmatic, but it leaves an obvious question hanging. A code negotiated with the very companies it governs risks codifying what they were already willing to do, rather than what oversight demands.

Industry has pushed back on the pace as much as the substance. Several large developers argue that obligations on transparency, copyright and systemic-risk testing arrived before the technical standards needed to meet them, leaving compliance teams to guess at moving targets. European firms, meanwhile, complain that the heaviest burdens fall on those least able to absorb them, and warn that the continent is writing rules for a technology it does not yet build at scale. That tension, between governing AI and growing it, runs through every Brussels debate on the file.

The Commission has signalled some sympathy. Talk of a broader simplification package, trimming overlapping digital reporting duties, has fuelled speculation that parts of the AI timetable could be softened or clarified. Supporters frame this as sensible housekeeping that removes duplication without touching core safeguards. Critics see the early sign of a retreat, a worry that the bloc will blink the moment its flagship law starts to bite. Both readings will be tested over the coming year.

What is no longer in doubt is that the AI Office must move from drafting to doing. It is hiring technical staff capable of auditing systems most regulators have never examined, building channels to receive incident reports, and preparing to wield fines that can reach a substantial share of global turnover. Whether it can match the resources and speed of the firms it supervises remains the open question. Europe has written the world’s most detailed rulebook for artificial intelligence. The harder task, proving it can be enforced, is only beginning, and the credibility of the entire experiment rests on getting that part right.

Comments