The EU Just Classified ChatGPT as a Search Engine. The Fine Is 6% of Global Revenue.

What happened

The European Commission designated ChatGPT, Reddit, and Roblox as "very large" services under the EU Digital Services Act, triggering the bloc's most stringent platform obligations. All three must be in full compliance by 31 December 2026; fines for non-compliance reach 6% of global revenue.

Context

The DSA has been in force since 2024. Services cross into the "very large" tier at 45 million average monthly EU users, and all three of these platforms have surpassed that line. The framework already covered social media and search, but this is the first time a generative-AI chatbot has been formally classified under it. X's AI chatbot Grok is separately under a DSA investigation, which suggests the Commission is treating AI-assisted content services as in-scope even when they have not yet been formally designated.

How it works

The "very large" classification triggers a defined set of obligations: removing illegal content, protecting minors' privacy and security, restricting ad targeting (no targeting of minors; no targeting by sexual orientation, religion, ethnicity, or political belief), and providing transparency about how recommendation algorithms work. OpenAI will be held accountable specifically for mitigating risks related to ChatGPT's impact on minors, user mental health, and the spread of illegal content in the EU.

One classification detail carries weight. The Verge reports the Commission designated ChatGPT as a "Very Large Online Search Engine" (VLOSE), a distinct DSA category, while Ars Technica describes all three services collectively as "very large online platforms." Whether VLOSE carries different obligations than VLOP is not clarified in either source. EU executive vice-president Henna Virkkunen described the designations as a higher standard of scrutiny and accountability.

Our read

The obvious reading is a regulatory checkbox for three large companies. The less obvious point is what classifying ChatGPT as a search engine actually means. The Commission is treating a generative-AI assistant that produces answers rather than ranked links as functionally equivalent to a search engine under the DSA. That is a legal classification with operational consequences: OpenAI must now explain how its ranking and recommendation systems work to EU regulators, and minor-protection and illegal-content mitigation become first-order engineering requirements rather than best-effort features.

The second-order effect lands on every team that builds on ChatGPT's API or consumer product in the EU. If OpenAI tightens output filtering, adds age-gating, or restricts what the model can discuss to satisfy its new obligations, downstream workflows shift. No source in the brief details what specific technical changes are required, and the scope of algorithmic-transparency disclosures a VLOSE must make versus a VLOP is unresolved. What is stated is the obligation and the deadline; the implementation is still open.

The 6%-of-global-revenue fine changes the risk calculus in a way that matters. For OpenAI that is a nine-figure sum. It makes compliance a board-level priority, which means the changes will likely be conservative, visible, and broader than the minimum the DSA text requires.

What this changes

For a small studio running ComfyUI, local open-weight models, and self-hosted inference, nothing changes on Monday. The 45-million-user threshold is not in question, the studio is not a DSA-regulated entity, and the 6% fine applies to the platform, not to content creators or downstream users.

The indirect items to watch between now and December 2026: if the workflow calls the ChatGPT API for scripting or prompt work and serves EU clients, OpenAI may alter output filtering, add disclaimers, or change availability. If finished content is distributed on Reddit, that platform's new illegal-content and minor-protection duties could change moderation outcomes for AI-generated clips. Neither source states what those changes will look like. The honest answer is: monitor, do not re-architect yet.

License

This is a regulatory designation, not a software or model release. No source in the brief attaches a software or content licence to any component of the DSA classification. No licence applies.

Key takeaways

• The EU has formally classified a generative-AI chatbot under the DSA, extending the framework into AI-assisted content for the first time.
• ChatGPT, Reddit, and Roblox must comply by 31 December 2026; fines for non-compliance reach 6% of global revenue.
• The Verge calls ChatGPT a VLOSE; Ars Technica groups all three as VLOPs. The sources do not clarify whether the distinction changes obligations.
• No source details the specific technical changes the three services must implement to comply.
• Small studios and downstream API users are not directly regulated but should monitor for behavioural changes in ChatGPT's EU-facing product and API.

Sources

1. ChatGPT to face tougher regulation in the EU — tier 2
2. ChatGPT and Reddit now face EU's toughest online safety rules — tier 2