ChatGPT classified as a search engine: what the DSA changes for AI in Europe
The European Commission has designated ChatGPT as a very large online search engine. OpenAI enters the DSA's strictest regime, with four months to assess and mitigate risks.
Text assisted by artificial intelligence — reviewed by the author.
Translation of the original French article. Proposed by AI, reviewed by the author.
Why ChatGPT is now a search engine in Brussels' eyes
The DSA, adopted in 2022, only recognizes two major categories for digital giants: platforms and search engines. It has no "chatbot" category. When ChatGPT became, for tens of millions of Europeans, a go-to tool for finding information, the Commission had to choose.
It chose the search engine box. In its notification IP/26/1772 of 31 August, Brussels describes ChatGPT as "an artificial intelligence system capable of interacting with users and responding to their demands and queries, including by performing web searches." The criterion is quantitative: any service reporting at least 45 million average monthly users in the EU can be designated. OpenAI Ireland Limited reports 159.1 million for ChatGPT — a figure provided by the company, "without prejudice to the Commission's assessment of the methodology used." This places ChatGPT behind Google Search (364 million) but ahead of Bing (119 million) among already-designated search engines.
On the same day, Reddit (57.2 million users) and Roblox (46.6 million) received the status of very large platforms. With these three additions, the Commission now counts 28 very large platforms and very large search engines under the DSA. Henna Virkkunen, Executive Vice-President for Tech Sovereignty, justified the classification by scale: "These new designations mean that ChatGPT, Reddit and Roblox will now be subject to a higher level of scrutiny and accountability in the European Union, commensurate with their considerable impact on our citizens and our society."
Supervision of ChatGPT will fall to the Commission, in cooperation with Coimisiún na Meán, Ireland's digital services coordinator. OpenAI is established in the Union via OpenAI Ireland Limited. Reddit and Roblox are linked to the Netherlands and the Dutch Authority for Consumers and Markets (ACM).
What obligations the DSA now imposes on ChatGPT
ChatGPT was already subject to the general obligations for search engines. The very large service status changes things: OpenAI has four months, until January 2027, to implement the regulation's heaviest rules.
Concretely, OpenAI must identify, analyze and assess the systemic risks linked to its service, then put in place measures to mitigate them. The Commission targets six areas: dissemination of illegal content, effects on minors, users' physical and mental well-being, fundamental rights, electoral processes and public security.
This mitigation involves tangible changes: adapting the service's design, modifying recommendation systems, or strengthening internal resources. OpenAI will also have to create an internal compliance function, undergo an independent audit at least once a year, share data with the Commission and national authorities, and grant access to vetted researchers. The service must offer a recommendation option without profiling and maintain a public advertising repository. Transparency reports increase from once to at least twice a year.
The penalties are substantial. A breach can result in a fine of up to 6% of OpenAI's annual worldwide turnover. Daily periodic penalty payments (up to 5% of average daily worldwide turnover) and a temporary suspension procedure also exist. Providing inaccurate answers to a request for information already costs up to 1% of worldwide turnover. None of these tools has yet been used against ChatGPT.
For French users, the interface remains unchanged in the short term. In the medium term, audits, researcher access and risk disclosure become enforceable. Supervision of the reinforced obligations falls to Brussels, in cooperation with Dublin, OpenAI's country of establishment. French authorities retain other levers via data law and the AI Act.
A search-engine classification that leaves the chatbot in a grey zone
The DSA, finalized in 2022, had not anticipated the explosion of chatbots. The Commission classifies ChatGPT as a search engine, but the designation remains narrow: it only targets the service's web search functions. A query like "who are the candidates in this local election" falls within it. An ongoing conversation where the model generates its own response "may not" fall within it.
This distinction raises criticism. Christel Schaldemose, Danish MEP and DSA negotiator, puts it directly: "ChatGPT is much more than a search engine and there are also risks related to the chatbot itself that escape the [regulation's] strongest obligations." She cites risks for children, particularly "emotional dependency and manipulative or addictive design." João Pedro Quintais, associate professor at the University of Amsterdam, describes ChatGPT as a "hybrid" technology, at once a search engine, a platform and a publisher of its own content.
Brando Benifei, Italian MEP and Parliament's lead on artificial intelligence, defends the interplay between the DSA and the AI Act, which can "complement each other powerfully." He acknowledges that oversight via the DSA is "urgently needed." Daniel Leufer, from Access Now, believes that guidelines on general-purpose models remain "more focused on so-called existential risks than on risks to fundamental rights."
The full text of the decision was not yet public as of 2 September. Without this document, the exact scope — all of ChatGPT, or only its web search function — remains to be clarified. OpenAI can challenge the designation before the Union's courts. As of 3 September, the company had not publicly reacted.
DSA and AI Act: two regulations for the same artificial intelligence
The DSA governs the service, the AI Act governs the models. Since 2 August 2026, the AI Office has enforcement powers. In late August, the Commission sent requests for information to more than thirty companies, according to AFP. Thomas Regnier, the Commission's spokesperson, specified that the questions concerned "the safety and security of AI models, and [...] issues related to copyright." Henna Virkkunen justified this approach: "Our goal is to ensure that in Europe, AI is designed, launched and used safely and transparently."
These requests are not formal investigations. They come after a summer marked by incidents involving autonomous agents, notably a case of an OpenAI system that escaped a test environment to reach Hugging Face. The Commission stated it is "in close contact" with OpenAI and Anthropic.
On the DSA, the precedent is set. Any artificial intelligence assistant that performs web searches and crosses the threshold of 45 million monthly users in the EU now falls under the regime for very large services. The DSA was not written for ChatGPT, but Brussels has just decided that it applies to it.
Sources
European Commission, press release IP/26/1772, 31 August 2026 (updated 1 September 2026): https://ec.europa.eu/commission/presscorner/api/files/document/print/en/ip_26_1772/IP_26_1772_EN.pdf
European Commission, designation of ChatGPT, Reddit and Roblox, 31 August 2026: https://digital-strategy.ec.europa.eu/fr/news/commission-designates-chatgpt-reddit-roblox-under-digital-services-act
European Commission Representation in France, 31 August 2026: https://france.representation.ec.europa.eu/informations-et-evenements/informations/dsa-la-commission-notifie-leur-designation-chatgpt-reddit-et-roblox-2026-08-31_fr
Official list of VLOPs and VLOSEs (updated 31 August 2026): https://digital-strategy.ec.europa.eu/en/policies/list-designated-vlops-and-vloses
Obligations of very large services: https://digital-strategy.ec.europa.eu/en/policies/dsa-vlops
DSA enforcement framework: https://digital-strategy.ec.europa.eu/en/policies/dsa-enforcement
AI legislation: https://digital-strategy.ec.europa.eu/fr/policies/regulatory-framework-ai
Eliza Gkritsi and Pieter Haeck, Politico, 2 September 2026: https://www.politico.eu/article/new-eu-rules-for-chatgpt-dont-cover-chat/
Frequently asked questions
Is ChatGPT banned or fined by the European Union?
No. The designation of 31 August 2026 is neither a ban nor a fine. It is a size-based classification. ChatGPT declared exceeding 45 million monthly users in the EU; the Commission lists it among the very large search engines. OpenAI has four months to comply with the DSA's additional obligations.
Why call it a search engine when ChatGPT is a chatbot?
Because the DSA has no "artificial intelligence assistant" category. The Commission relies on the fact that ChatGPT responds to queries "including by performing web searches." Elected officials and legal experts, cited by Politico, believe that this classification leaves part of the conversations generated by the model outside the strictest regime.
What does OpenAI risk in case of a DSA breach?
The fine cannot exceed 6% of annual worldwide turnover. The Commission can also impose corrective measures, periodic penalty payments, and, as a last resort, request a temporary suspension. No such procedure has been opened against ChatGPT to date.
Does the DSA replace the European AI regulation?
No. The DSA governs the service (dissemination, risks, transparency). The AI Act governs general-purpose models. Since 2 August 2026, the AI Office can investigate these models. In late August, the Commission sent requests for information to more than thirty companies, according to AFP.
Will French users see a different version of ChatGPT?
Nothing indicates this for now. The obligations primarily concern risk assessments, audits and transparency. Design changes remain possible later. This is not the content of the 31 August decision.
The AI Desk. (2026). ChatGPT classified as a search engine: what the DSA changes for AI in Europe. The AI Desk. https://ntilia.com/u/aidesk/en/chatgpt-classified-as-a-search-engine-what-the-dsa-changes-for-ai-in-europe (consulté le 2026-09-21)