Regulators shift to implementation
Countries are moving from headline AI laws to the gritty work of enforcement and infrastructure — Britain’s AI minister publicly criticized OpenAI after it paused a major 'Stargate' data‑centre project. Meanwhile, Europe is already hashing out technical compliance, with analysis pointing to specific logging requirements and deadlines under the EU AI Act. (UK AI Minister Criticizes OpenAI Over Halted Stargate Data Center Project - Bloomberg, What the EU AI Act requires for AI agent logging - Help Net Security)
Britain’s AI minister publicly rebuked OpenAI after the company paused a planned U.K. data-center buildout, shifting the AI fight from slogans to power, permits and rules. (bloomberg.com) At a Bloomberg event in London on April 16, Kanishka Narayan said “nothing has changed” in the site’s energy prices or regulatory treatment since OpenAI first committed to the project. OpenAI had blamed U.K. energy costs and regulation for putting “Stargate U.K.” on hold. (bloomberg.com) OpenAI said on April 9 that it would move forward only when “the right conditions” support long-term infrastructure investment. The company had said in September it would explore leasing up to 8,000 Nvidia graphics processors from Nscale in early 2026, with potential to scale to 31,000 over time. (cnbc.com) The U.K. government spent the past year pitching data centers as the “engines of the AI age.” Its AI Opportunities Action Plan, published January 13, 2025, called for faster buildout of compute capacity and planning reforms to make new facilities easier to build. (gov.uk) That push runs into the hard limits of the grid and the law. Politico reported the paused sites included Cobalt Park in northeast England, part of a government-designated “AI Growth Zone,” while OpenAI also pointed to predictable copyright rules as a factor in investment decisions. (politico.eu) In Europe, the argument has already moved from whether to regulate AI to how to prove a system did what it did. Under the European Union’s AI Act, high-risk systems must be able to record events automatically over the system’s lifetime. (helpnetsecurity.com) That logging rule hits systems used for credit scoring, hiring, insurance pricing, healthcare benefits and emergency-call triage, according to a compliance analysis published April 16. The article points to Articles 12, 13, 19 and 26, which together require automatic logs, instructions for deployers and at least six months of log retention. (helpnetsecurity.com) The deadlines are no longer abstract. The European Commission says the AI Act entered into force on August 1, 2024; rules on prohibited practices started applying on February 2, 2025; obligations for general-purpose AI models started on August 2, 2025; and the Commission’s enforcement powers for those models apply from August 2, 2026. (commission.europa.eu, digital-strategy.ec.europa.eu) The same Commission guidance says providers of general-purpose AI models placed on the market after August 2, 2025 must comply now, while providers of models already on the market before that date have until August 2, 2027. Most high-risk system obligations also begin applying on August 2, 2026. (digital-strategy.ec.europa.eu, op.europa.eu) OpenAI said it still sees “huge potential” in the U.K., is expanding its London research presence and remains in discussions with Nscale. Regulators, meanwhile, are moving ahead on the less visible work: electricity supply on one side, audit trails on the other. (politico.eu, helpnetsecurity.com)
