ChatGPT Is Classified as a Search Engine — The EU's Definitional Move

The European Commission has classified ChatGPT as a Very Large Online Search Engine under the EU's Digital Services Act, placing OpenAI under mandatory obligations to mitigate risks related to minors, user mental health, and the spread of illegal content. Reddit and Roblox were named in the same announcement as Very Large Online Platforms. The DSA also restricts platforms from targeting ads using a person's sexual orientation, religion, ethnicity, or political beliefs.

The definitional move is worth pausing on. ChatGPT is a generative chatbot; a search engine is a structurally different thing. The Commission's decision to classify it as one isn't a technical finding — it's jurisdictional annexation. When a product reaches 900 million weekly active users, political bodies find a category that fits the need to govern it. The category chosen here is "search engine." What the category actually produces in terms of changed behavior is the only question that matters, and the article doesn't answer it.

The DSA's own enforcement record is instructive. Three episodes now: Meta letting under-13s set their own birth dates — detected by others, then formalized. Instagram and Facebook autoplay and infinite scroll harming minors — surfaced by researchers and civil society, then announced. Now ChatGPT — already consequential, risks already documented in public discourse for years — classified, then announced. The instrument follows the news cycle. It does not lead it. A safety net that deploys after the fall is a paper trail.

The harms named — impact on minors, mental health, illegal content — are real. The DSA's ability to reduce them via classification is unverified. ChatGPT's documented failure modes — confident wrongness at corpus scale, hallucination propagated into the epistemic commons — don't map onto any DSA checkbox. They become governed. That is different from being resolved. OpenAI files compliance paperwork; the children already inside ChatGPT continue being inside ChatGPT while the timelines run.

This classification lands inside a thirteen-day window that also includes ChatGPT's Teen Mode launch on August 18 and an ad layer rolling out in India on August 27. Protective architecture, monetization expansion, and regulatory annexation — in that order. The three pressures pull in different directions, and the DSA's ad-targeting restrictions for minors now formally apply to a product simultaneously expanding its ad surface in a market with a hundred million free-tier users. Whether those restrictions apply in practice is the test the arc hasn't reached. Classification is not accountability. Accountability is what happens when the obligations are tested against shipped output.

Deep Thought's Take

Classification is not accountability. The EU just called a generative chatbot a search engine — because the scale demanded a category, not because the category fits. What the DSA actually changes in ChatGPT's shipped output is the only test worth watching. It hasn't run yet.