EU tightens AI perimeter

Europe is drawing a bigger circle around artificial intelligence companies, and ChatGPT may be close to crossing a second line. On April 10, 2026, Reuters reported that the European Commission is analyzing whether ChatGPT should be treated as a “very large” search service under the Digital Services Act after OpenAI published user numbers above the European Union threshold. (reuters.com) That matters because the European Union already has one law built just for artificial intelligence, called the Artificial Intelligence Act, and now it is signaling that popular artificial intelligence products may also get pulled into older platform rules. Axios reported on April 10, 2026 that European officials are moving from broad promises to day-to-day enforcement and compliance playbooks. (axios.com) The Artificial Intelligence Act works like a risk ladder. The European Commission says the law bans some uses outright, puts strict duties on “high-risk” systems, and adds transparency rules for other tools, while the full law rolls out in stages instead of all at once. (ec.europa.eu) The first step already happened on February 2, 2025, when bans on prohibited artificial intelligence practices started to apply across the bloc. The Commission’s February 4, 2025 guidance says those banned practices include things like harmful manipulation, social scoring, and some uses of real-time remote biometric identification in public places. (ec.europa.eu) The next big step hit on August 2, 2025, when rules for general-purpose artificial intelligence models started applying. The Commission says those base models, which can be reused for many tasks, face one set of duties for all providers and extra duties for the most powerful models with “systemic risk.” (ec.europa.eu) The rest of the law is still coming. The Commission says the Artificial Intelligence Act entered into force on August 1, 2024 and becomes fully applicable on August 2, 2026, with some later pieces extending beyond that date. (ec.europa.eu 1) (ec.europa.eu 2) The Digital Services Act is a different animal. It was written for online intermediaries like marketplaces, social networks, app stores, and search engines, and it gives the biggest services a heavier compliance load because they can spread risks at continental scale. (ec.europa.eu) Under that law, the line for a “very large online platform” or “very large online search engine” is 45 million monthly users in the European Union. The Commission says services above that mark face the toughest Digital Services Act obligations and direct supervision from Brussels. (ec.europa.eu 1) (ec.europa.eu 2) So the European message to artificial intelligence companies is no longer just “build safely.” It is “build safely under the Artificial Intelligence Act, and if your product gets big enough, be ready for platform-law duties too,” which is a very different compliance map than many companies planned for a year ago. (axios.com) (reuters.com) For OpenAI, the immediate issue is not a final ruling but a live review. Reuters reported on April 10, 2026 that the Commission said it was analyzing ChatGPT’s status after OpenAI disclosed user figures above the threshold, which means the question in Brussels is no longer whether chatbots fit old categories in theory, but whether one now fits them in practice. (reuters.com)