EU Designates ChatGPT as a Very Large Online Search Engine Under the DSA

On August 31, 2026, the European Commission crossed a regulatory Rubicon: per Bloomberg’s report, ChatGPT was formally designated a Very Large Online Search Engine (VLOSE) under the Digital Services Act, alongside Reddit and Roblox as very large online platforms. It is the first time a generative AI assistant has been brought under the DSA’s strictest tier, and the classification itself is the interesting part: the EU treated ChatGPT as a search engine, not a social network.

Why the Search Engine Label Matters

The VLOSE tier applies to services with more than 45 million average monthly EU users, and ChatGPT is far past it. OpenAI’s own DSA transparency disclosure put ChatGPT search at roughly 159 million average monthly EU users for the six months ending March 2026. But the category choice determines which obligations bite. As a VLOSE, ChatGPT must run systemic risk assessments covering, among other things, election integrity and minors’ safety, open data access to vetted researchers, submit to independent audits, and explain how its recommendation and ranking systems work. For a product where “ranking” means model behavior and retrieval grounding, those obligations translate directly into model-level transparency demands, not just feed-tuning the way they do for social platforms.

OpenAI has four months to demonstrate compliance from designation, with the threat of fines up to 6% of global annual revenue for serious breaches. Reddit and Roblox, designated the same day, face the familiar social-media obligations; ChatGPT is the experiment in applying them to a model product.

The Precedent Problem for AI Builders

The immediate impact lands on OpenAI, but the precedent lands on everyone shipping AI systems in Europe. Designation was triggered by user numbers, and any assistant or AI search product that crosses the threshold inherits the same playbook: systemic risk evaluations, adversarial testing evidence, researcher API access, and algorithmic disclosure. Teams building on top of foundation models should also note that DSA duties reach the service layer, so a product wrapping GPT models in the EU is not insulated by pointing at its provider.

This lands in the same month the EU AI Act’s transparency duties began applying to chatbots and synthetic content, meaning AI products in Europe now answer to two overlapping regimes: the AI Act for the model behavior, the DSA for the service behavior. Compliance teams that treated those as one workstream are about to discover they are two.