The EU Just Classified ChatGPT as a Very Large Online Search Engine

On 31 August 2026 the European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act (DSA), the first time an AI chatbot has been pulled into the EU’s strictest tier of platform rules. Reddit and Roblox were named Very Large Online Platforms on the same day. OpenAI now has four months — until roughly January 2027 — to meet a long list of obligations covering systemic-risk assessments, independent annual audits, data access for regulators and researchers, and stronger complaint and appeal mechanisms. Ignoring those duties can cost up to 6% of a company’s global annual turnover.

I have been writing about AI regulation on this site for more than a year, and this is the moment the framing finally shifted in public. Brussels is no longer treating a chatbot purely as an “AI model” to be handled by the AI Act. It is treating ChatGPT as mass-market infrastructure for finding information — the same category as a search engine — with all the societal-risk baggage that label carries.

What actually happened

The Commission’s reasoning hinges on one detail: ChatGPT can search the live web and return answers with links. That makes it, in the Commission’s words, a hybrid service that qualifies as an online search engine under the DSA. Because OpenAI itself reported about 159.1 million average monthly users in the EU — more than three times the 45-million threshold that triggers “very large” status — the designation followed almost automatically.

Reddit (57.2 million EU users) and Roblox (46.6 million) were designated Very Large Online Platforms in the same announcement. Commission Vice-President Henna Virkkunen put it plainly: “These new designations mean that ChatGPT, Reddit and Roblox will now be held to a higher standard of scrutiny and accountability in the European Union, in line with their large impact on our citizens and society.”

OpenAI’s response was careful. A spokesperson narrowed the scope to “ChatGPT search,” saying that feature “operates as a search service under the DSA” — a framing that is noticeably tighter than the Commission’s blanket description of the product. That gap is going to matter over the next year, because it decides how much of ChatGPT’s behaviour gets audited.

The obligations that come with the label

Being a VLOSE is not a badge. It is a compliance program. Within four months OpenAI is expected to:

• Run formal systemic-risk assessments covering illegal content, effects on minors, physical and mental health, fundamental rights, civic discourse, elections and public security — and show how its algorithms feed those risks.
• Submit to an independent external audit at least once a year and respond, on the record, to the auditor’s recommendations.
• Give the Commission and national regulators access to the data needed to monitor compliance, and set up a vetted-researcher data-access channel.
• Provide clear notice-and-action tools, an internal complaint system, out-of-court dispute resolution, and a “trusted flagger” pipeline for priority reports.
• Publish transparency reporting and keep an ads repository if it runs advertising.

Enforcement sits directly with the Commission rather than a single member state, coordinated with Ireland’s Coimisiún na Meán because OpenAI’s EU base is in Dublin. The maximum penalty under the DSA is 6% of worldwide annual turnover, with periodic penalty payments on top for continued non-compliance.

Quick reference

Why the “search engine” framing is the real story

There has been a long argument in EU legal circles about whether a chatbot that answers questions is functionally a search engine. The Commission has now answered it in practice. If a service is where tens of millions of people go to ask “what happened” or “is this true,” then it shapes public understanding the way Google Search or Bing does, and it should carry the same duties around misinformation, election integrity and manipulated media.

That has consequences beyond OpenAI. Google’s Gemini, Microsoft Copilot, Perplexity and Meta AI all blend generation with web retrieval. Any of them that clears 45 million EU users on a comparable feature is now a plausible next candidate. The EU has effectively created a template: cross the user line, gain the ability to retrieve and summarise the open web, and you inherit search-engine-grade accountability.

How this stacks with the AI Act

The DSA designation does not replace the EU AI Act — it runs alongside it. OpenAI already has general-purpose-AI obligations under the AI Act (documentation, copyright policy, systemic-risk evaluation for the most capable models). The DSA adds a second, overlapping regime focused on content risk, transparency and user redress. Companies operating at this scale in Europe are now managing dual regulatory tracks, and the compliance teams for each are not the same people.

This is the pattern I keep coming back to in coverage of the EU’s order forcing Android to open up to rival AI assistants and of Colorado’s state-level AI law: regulators are no longer waiting for one comprehensive AI statute to do everything. They are reaching for whatever existing framework fits — competition law, consumer law, platform law — and applying it now.

What changes for people using ChatGPT

In the near term, not much that is visible. The obligations are mostly structural — audits, risk documentation, regulator access. Over the next year EU users should expect clearer reporting mechanisms inside the product, more transparency about how answers are ranked and sourced, published risk assessments, and possibly changes to how ChatGPT handles topics tied to elections, health and minors. If OpenAI and the Commission end up disagreeing about scope, that fight will play out in formal proceedings rather than in the app.

Frequently asked questions

Is ChatGPT banned or restricted in the EU now?

No. The designation adds compliance duties; it does not limit access. ChatGPT continues to operate normally across the EU while OpenAI works toward the January 2027 deadline.

Why is ChatGPT called a “search engine” and not an “AI platform”?

Because it can query the live web and return sourced answers. Under the DSA that capability, combined with its scale, makes it a hybrid service that meets the legal definition of an online search engine, which is why it was designated a VLOSE rather than a VLOP.

What is the difference between a VLOSE and a VLOP?

A VLOP (Very Large Online Platform) is a large hosting or social service — Reddit and Roblox fall here. A VLOSE (Very Large Online Search Engine) is a large search service. The obligations overlap heavily, but search engines have specific duties around result ranking and query-based risk.

Could Gemini or Copilot be designated next?

Potentially. The Commission has signalled it views widely used AI assistants that retrieve web content as candidates. Any that self-report or are found to exceed 45 million average monthly EU users on a comparable feature could face the same process.

How large is the possible fine?

Up to 6% of a company’s total worldwide annual turnover for serious or repeated breaches, plus recurring penalty payments while non-compliance continues. For a company OpenAI’s size that is a material number.

When do the new rules take full effect?

OpenAI has four months from notification, putting the compliance deadline in roughly January 2027. The first independent audit cycle would follow after that.

The bottom line

The EU has decided that once an AI assistant becomes one of the main places people go to find things out, it should be regulated like the front page of the internet — audited, risk-assessed and answerable to a regulator. ChatGPT is the first to cross that line officially. It will not be the last, and the next twelve months of audits and scope arguments will set the tone for how the largest AI products operate in Europe.

Primary source: the European Commission’s designation announcement.