Imagine an athletics federation announcing world leadership in sprinting. Its runners are talented. Its equipment is excellent. Unfortunately, the starting pistol requires an interoperability assessment, the floodlights cost twice as much to operate, and the committee is still consulting on whether running infringes someone’s intellectual property.
Welcome to the less inspiring side of European AI ambition. The EU and UK have genuine research and industrial strengths. They also offer a masterclass in making access, investment and policy clarity harder than the speeches suggest. This is not one grand conspiracy against technology. It is a collection of expensive obstacles wearing a leadership badge.
World-class waiting rooms
The November 9 snapshot of the independent “AI features delayed in the EU” tracker lists 16 unavailable items, including one discontinued product, six partly available and seven that arrived late. Its 40-day average applies to those seven late arrivals, not every missing feature. That distinction matters: a waiting-room clock is not a delivery guarantee.
Nor does the list establish that regulation caused every delay. Companies make rollout decisions for multiple reasons. But users experience the outcome, not the explanatory slide deck: a launch announcement followed by the familiar discovery that their location remains an experimental incompatibility.
Apple supplies a concrete regulatory dispute. It says the Digital Markets Act prevents Siri AI shipping on EU iPhones and iPads with iOS 27 and iPadOS 27, and gives no availability timeline. Its account blames disagreements over access for rival assistants and privacy safeguards. That is Apple’s position, not an impartial judgment on the dispute. Siri AI remains available on macOS 27 and visionOS 27 in the EU, so this is not an Apple-wide blockade.
For the customer, the allegory is still magnificent: two officials argue over the safest design for a bridge while the commuter is instructed to admire the river.
The timetable that needs its own timetable
According to Ferraz & Whitmore’s legal analysis, the Digital Omnibus moved Annex III high-risk AI obligations to December 2, 2027. Product-embedded Annex I systems retain their August 2, 2028 date; that date was not newly postponed. Meanwhile, transparency requirements and the AI Office’s enforcement regime took effect on August 2, 2026.
Calling this a blanket regulatory pause would therefore be wrong. It is a selective extension inside an already complicated framework. Brussels has not abolished the obstacle course. It has moved several hurdles and issued an updated map.
ASML chief Christophe Fouquet put the industrial concern bluntly in his Politico interview: “If they cannot do it here, they will do it elsewhere.” The Commission’s response is that the AI Act builds trust, encouraging adoption and investment. Trust matters. So does ensuring there is something available to trust.
The electricity bill joins the negotiating team
Rules are only part of the story. CNBC reports that Europe’s prices for energy-intensive industries were roughly double US levels last year, citing the International Energy Agency. Nvidia principal programme manager Vladimir Prodanovic offered the sharper verdict: “The middle part of Europe has already lost the game.” That is one executive’s assessment, not a mathematical proof of continental doom.
The geography matters. CNBC identifies the Nordics and France as potential beneficiaries of cheaper power, not fellow passengers on an inevitably sinking European ship. It also reports that OpenAI paused its UK Stargate project partly because of energy costs, with the regulatory environment another concern. A pause is not a confirmed permanent cancellation.
Data centres cannot run on ministerial enthusiasm. A sovereignty strategy without competitive power is a restaurant promising independence from foreign chefs while charging its own kitchen double for gas.
Britain consults, reverses and restricts
The UK’s March copyright report says a broad AI-training exception with an opt-out is no longer the government’s preferred approach. It promises more evidence-gathering and consideration of alternatives, not a newly enacted blanket training ban. Creators’ concerns about consent and compensation are real. So is the cost of leaving developers unsure which framework will eventually emerge.
Then there is Grok. Ofcom’s October update describes UK-IP restrictions on pornographic image content and sexually explicit companion-chatbot conversations, alongside an ongoing investigation into X’s illegal-content duties. It does not establish that Britain’s general-purpose Grok reasoning or ordinary answers have been broadly disabled. Turning that document into proof of universal political censorship would be satire outrunning its evidence.
The narrower criticism is enough: Britain advertises AI leadership while combining unresolved training policy, costly infrastructure and geographically restricted services. Europe wants technological sovereignty, but sovereignty requires building and using technology, not merely supervising its absence.
Safety, competition and creators’ rights deserve rules. They also deserve rules that work. Otherwise, the EU and UK risk becoming the world’s finest lighthouse inspectors: meticulous about every bulb, delighted with their paperwork, and increasingly puzzled that the ships are docking somewhere else.
Sources
- AI features delayed in the EU (Updated 9 Nov 2026)
- Due to DMA, Siri AI delayed in EU for iOS 27 and iPadOS 27
- EU AI Act Digital Omnibus: High-Risk Deadlines Moved to December 2027
- Europe’s top tech boss warns Brussels is driving AI companies away
- High energy prices could derail Europe’s AI race with U.S. and China
- Report on Copyright and Artificial Intelligence
- Ofcom update: Monitoring X’s response to hate and terror content, and Grok investigation latest