On 31 August 2026, the European Commission designated ChatGPT as a Very Large Online Search Engine, or VLOSE: meaning it treats it as a very large online search engine under the Digital Services Act, the EU's regulation on digital services.
The designation triggers enhanced oversight. OpenAI Ireland Limited, which provides the service in the Union, has four months from the notification to comply with the obligations arising from this decision. It will have to assess and mitigate systemic risks — those linked to the service operating at scale, including risks to minors and mental wellbeing — under the Commission's supervision.
To be designated as a very large search engine under the DSA, a service must reach at least 45 million average monthly active users in the EU. For ChatGPT search, OpenAI declared approximately 159.1 million for the half-year ending 31 March 2026. That figure covers the online search function, while in the Commission's list the designated service appears under the broader name ChatGPT.
The distinction matters because ChatGPT can both search the web and summarise pages, and generate responses without consulting the internet at all. If oversight applied only to search, some risks tied to conversations could fall outside its scope.
That is the concern of Christel Schaldemose, a Danish MEP who took part in negotiating the DSA. She told Politico there is a risk of users forming emotional attachments to the chatbot to the point of dependency, and of features that shape users' choices or nudge them to keep using it — particularly when those users are minors.
To know whether this gap actually exists, one would need to examine the reasoning behind the designation. The full text of the decision on ChatGPT is still not available in the Commission's list of services subject to enhanced oversight. Legal scholar João Pedro Quintais of the University of Amsterdam has also written that he is waiting for it to be published.
Daniel Leufer, head of emerging technology policy at Access Now, a digital rights organisation, welcomes the designation because it can make ChatGPT's workings and the risks to users' rights more transparent. He calls for attention to the filters applied to inputs and outputs: from refusing to engage with a topic to the way it gets reframed, these choices shape the information people receive.
Legal scholar Tobias Mast, of the Leibniz-Institut für Medienforschung in Hamburg, considers the choice to treat ChatGPT as a search engine defensible. The DSA's definition allows for results in any format: according to Mast, a conversational response can fall within it too. Many of the rules, however, were built for services that distribute other people's content and sit less comfortably with a chatbot that generates its own.
On X, Lazar Radic, a law professor at IE University and competition expert at the International Center for Law & Economics, criticises rules that trigger obligations based on a service's size. He invokes market power, the central concept in antitrust: the ability to distort competition. Under the DSA, the user threshold follows a different logic: a very widely used service can expose many people to the same risks, even without dominating the market.
OpenAI will now have to document the service's risks and the measures taken to reduce them. For Mast, these assessments will need to account for how people actually use ChatGPT, not just the results of laboratory tests.