Brussels just handed OpenAI a new regulatory headache. The EU Commission classified ChatGPT as a very large search engine on Monday, pulling the chatbot under the Digital Services Act’s strictest oversight tier. The trigger? ChatGPT’s built-in web search combined with at least 45 million monthly EU users — the exact threshold that flips the switch. Reddit and Roblox got hit with the same reclassification, though as very large online platforms, since both let users share content from other users.

The three services now have until the end of December 2026 to comply. That’s four months to stand up risk assessments covering illegal content, dangers to minors, and election interference, plus a publicly viewable ad archive, semiannual transparency reports, a complaints process, a crisis response mechanism, and — most consequentially — data access for approved researchers. Henna Virkkunen, the Commission’s vice president, said all three would face stricter oversight. ChatGPT joins 27 other platforms and search engines already classified under the DSA.

Here’s where things get legally interesting. The researcher data access requirement is genuinely contested territory. Natali Helberger, a law professor at the University of Amsterdam, told Tech Policy Press the classification “will raise interesting questions regarding the extent of data access rights, and whether the access obligation under Article 40 could potentially also include access to training data or model weights if such access is necessary to identify systemic risk or assess mitigation measures.” The Commission, notably, doesn’t explicitly get the power to test ChatGPT’s models directly.

What’s striking here is how quickly the EU’s regulatory machinery has adapted to generative AI. When the DSA was drafted, the very large online platform category was built with social networks and marketplaces in mind — not chatbots that moonlight as search engines. Brussels has effectively said: if it walks like a search engine and reaches 45 million Europeans like a search engine, it gets regulated like one. The four-month runway is tight by policy standards, and OpenAI’s compliance team is likely already mapping how much model transparency they can give without exposing proprietary weights. That tension — between regulatory access and trade secrets — is where the real fight will happen.