EU eyes DSA for ChatGPT

Brussels is testing whether ChatGPT should be treated less like a software tool and more like a giant platform, because the European Union’s Digital Services Act kicks in at 45 million monthly users in Europe. The European Commission says services above that line can be labeled “very large online platforms” and moved into its toughest oversight tier. (digital-strategy.ec.europa.eu) That label was built for the biggest internet venues, the places where millions of people post, search, buy, and share every day. Once a service is designated, the Commission says it gets four months to comply with extra duties aimed at “systemic risks” that can spread across society at scale. (digital-strategy.ec.europa.eu) Those duties go beyond taking down illegal material. The Commission says very large online platforms have to assess risks tied to illegal content, fundamental rights, public security, elections, public health, minors, and mental wellbeing, then report those risks and put mitigation measures in place. (digital-strategy.ec.europa.eu) OpenAI has already been publishing Digital Services Act material in Europe through OpenAI Ireland Limited, which shows ChatGPT is not new to this rulebook. OpenAI’s help page says it has a designated contact point for European authorities under Article 11 of the law, and its 2025 transparency reporting package says the company is filing under Articles 15 and 24 of the Digital Services Act. (help.openai.com) (cdn.openai.com) The twist is that ChatGPT is also already inside a second European rulebook: the Artificial Intelligence Act. The European Commission says the Artificial Intelligence Act’s rules for general-purpose artificial intelligence models started applying on August 2, 2025, with enforcement powers for those model rules starting on August 2, 2026. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) That Artificial Intelligence Act framework looks at the model itself, the engine under the hood. The Commission says providers of general-purpose artificial intelligence models must prepare technical documentation, follow copyright rules, and publish a summary of training content, while the most powerful models face added duties including risk mitigation, incident reporting, and cybersecurity protections. (digital-strategy.ec.europa.eu) The Digital Services Act looks at something different: the service people actually use. That means Brussels is exploring a two-layer approach where one law governs the model and another governs the consumer-facing product that distributes answers, search results, and possibly harmful outputs to tens of millions of Europeans. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) If the Commission decides ChatGPT clears the threshold and fits the category, OpenAI would not just be answering questions about how its model was trained. It could also be asked how the product ranks or presents information, how it handles illegal or harmful material at scale, and how it measures broader social risks across the European Union. (digital-strategy.ec.europa.eu) (eur-lex.europa.eu) That would be a notable shift in how Europe sees chatbots. For years, regulators mostly treated them as artificial intelligence systems; this move would test whether a chatbot with mass reach should also be supervised like the biggest social networks and search engines. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) The immediate question is not whether Europe can regulate ChatGPT at all, because it already can under the Artificial Intelligence Act and existing Digital Services Act reporting rules. The question is whether Brussels now believes ChatGPT has become big enough, and influential enough, to fall into the same bucket as the internet’s largest public gateways. (digital-strategy.ec.europa.eu) (help.openai.com)