EU rewrites AI Act rules

The European Commission published draft guidelines on May 19 to clarify when artificial intelligence systems should be classified as “high-risk” under Article 6 of the EU AI Act. The move comes as Brussels reworks parts of the law after complaints from European technology companies that the rules were too difficult to apply in practice. Sifted reported this week that officials are revisiting elements of the framework after lobbying from some of the region’s fastest-growing firms. ### What exactly did Brussels put out this week? The Commission said the draft guidelines are meant to help providers, deployers and market-surveillance authorities decide whether a system falls into the AI Act’s high-risk category. That label matters because it triggers the law’s toughest compliance duties. (digital-strategy.ec.europa.eu) Article 6 sets out two main routes into the high-risk category. One covers AI used as a safety component, or as a product itself, under EU laws listed in Annex I, where the product must undergo third-party conformity assessment. The other route covers certain uses listed in Annex III. ### Why is Article 6 getting this much attention? (digital-strategy.ec.europa.eu) The Commission’s own summary says the guidance is intended to support “uniform application and effective enforcement” of Article 6. In practice, the classification question determines which systems face the heaviest documentation, testing and oversight burdens under the AI Act. (ai-act-service-desk.ec.europa.eu) Sifted reported that the broader law is being reconsidered after pushback from startups and other companies that argued parts of the regime were unworkable. The publication described the latest changes as a rethink driven by outcry from fast-growing European firms. ### Why are reliability concerns resurfacing at the same time? TechRadar reported on May 22 that a NewsBench study found major AI chatbots gave incorrect election information about 90% of the time. (digital-strategy.ec.europa.eu) The outlet said the findings showed repeated failures on election-related questions, a category where errors can carry obvious political consequences. (sifted.eu) The study does not determine how the EU will classify systems under Article 6, but it adds pressure to a debate already focused on which uses deserve the strictest obligations. The Commission’s guidance is about legal classification; the chatbot findings have sharpened a separate argument over whether current systems are reliable enough for politically sensitive tasks. (techradar.com) ### Why is the Vatican now part of the AI conversation? Religion News Service reported on May 22 that Pope Leo XIV’s new encyclical on artificial intelligence is due to be released on Monday, May 27, with Anthropic co-founder Chris Olah appearing alongside him. The report described the collaboration as an unusual link between the Vatican and a leading AI company. (digital-strategy.ec.europa.eu) That development does not affect the EU rulemaking process directly, but it shows how the argument over AI has widened beyond regulators and companies. The questions now being raised include not only compliance and product design, but also human dignity, ethics and the role of powerful model developers. (religionnews.com) ### What happens next in the EU process? The Commission said the consultation on the draft high-risk guidelines opened on May 19 and runs through June 23, 2026. The materials, including general principles and separate documents covering Annex I and Annex III, are posted on the EU’s digital-strategy site. (religionnews.com) June 23 is the next concrete deadline in this part of the process. Providers, deployers and national authorities can submit feedback before the Commission finalizes guidance on how Article 6 will be applied. (digital-strategy.ec.europa.eu)