The European Commission has classified ChatGPT as a very large search engine for the first time, putting the chatbot under stricter oversight rules. The move was announced in Brussels.
The reasoning is technical: OpenAI's chatbot can search the web and answer user queries. That makes it a search engine under the Digital Services Act. ChatGPT also reaches at least 45 million monthly users in the EU, the threshold for this classification.
In the same decision the Commission reclassified Reddit and Roblox as very large online platforms. Both count as platforms because users can share content from other users there.
Four months
The three services have until the end of December 2026 to meet the additional obligations. Chief among them is assessing risks to users: illegal content, dangers to minors and election interference.
The Commission gains wider audit powers as a result, exercised together with authorities in Ireland and the Netherlands. Commission vice president Henna Virkkunen said the services would have to submit to stricter oversight. That brings the number of platforms and search engines classified under the act to 28.
The concrete requirements
- A publicly viewable ad archive
- Data access for approved researchers
- Semiannual transparency reports
- Reporting mechanisms for illegal content, with a complaints process attached
- A crisis response mechanism
Where the law gets murky
The real argument is over the scope of data access. Whether the data opened to approved researchers must also cover training data or model weights is disputed among legal experts.
The difference is not trivial. Can systemic risk be assessed — measuring how a model behaves during an election period, say — from the outside by looking only at outputs, or does it require looking inside the model? The first is possible but limited; the second is the area companies guard most closely.
The question will set a precedent not only for ChatGPT but for every classification that follows.
The wider effect
The rule binds only inside the EU; the threshold is counted in EU users. In practice, though, the effect travels.
The reason: a company rarely maintains a separately engineered product for the regulated market. Transparency reports, content reporting mechanisms and crisis protocols get built into the product itself. A feature that is mandatory in the EU usually stays switched on everywhere, because running two versions is expensive.
So the line the EU draws tends to become a de facto global floor. The same thing happened with data protection rules.
Why this classification matters
The Digital Services Act was written with social networks and search engines in mind. A chatbot counting as a "search engine" shows how the law is being adapted to AI products: not by writing a new statute but by fitting them inside an existing definition.
The practical consequence: ChatGPT now sits in the same oversight frame as Google Search. The same road is open for any AI product that crosses the user threshold — and crossing it is easier today than it used to be.