The European Commission announced on Monday 1 September 2026 that ChatGPT will be classified as a very large online search engine under the Digital Services Act. The decision subjects OpenAI's chatbot to transparency and risk-mitigation obligations similar to those imposed on Google and Microsoft Bing, but only for features that function as search tools.

The classification marks the first time the EU has applied DSA rules to a generative AI chatbot. OpenAI now faces fines of up to 6% of its global annual revenue if it fails to comply with the new requirements. The move comes after growing public concern over AI-driven mental health advice and election-related misinformation.

Yet the ruling creates a significant regulatory gap. The Commission's narrow definition means conversational functions where the chatbot generates its own responses fall outside the DSA's strictest obligations. This distinction matters because users increasingly treat ChatGPT as a companion, therapist or political adviser rather than a search tool.

What the search engine classification covers

Under the Digital Services Act adopted in 2022, very large online search engines must implement transparency measures, conduct risk assessments and mitigate systemic risks. The threshold is 45 million monthly users in the European Union. ChatGPT clearly exceeds this threshold.

When a user asks ChatGPT to list candidates in a local election, that query now falls under DSA obligations. OpenAI must ensure the information is accurate and mitigate risks of misinformation. The same applies to search-like functions where the model retrieves and presents existing information from its training data.

The Commission took nearly one year to finalise this classification. Teams struggled to determine which category best fit a technology that did not exist when the DSA was drafted. The search engine designation was ultimately chosen over the very large online platform category, which typically applies to social networks and e-commerce sites.

Conversational functions remain outside strict oversight

The regulatory gap becomes apparent when users engage ChatGPT in extended dialogue. A conversation about which political candidate to support, where the chatbot offers opinions or analysis, may not trigger DSA obligations. Neither would therapeutic conversations where users seek emotional support or mental health guidance.

This distinction is not marginal. According to a recent study conducted by the insurance industry, 60% of adults worldwide use chatbots for therapeutic purposes. The use cases have drawn scrutiny after chatbots were linked to tragic suicides among teenagers, including the case of Adam Raine, a 16-year-old Californian whose parents have filed legal proceedings against OpenAI.

Christel Schaldemose, a Danish socialist MEP who was among the main negotiators of the DSA legislation, highlighted risks to children including emotional dependency and manipulative or addictive design. She urged the Commission to clarify how these risks are addressed under current regulation.

Why the Commission chose search engine over platform

The platform designation would have imposed additional content moderation obligations but might have allowed OpenAI to escape significant liability. The DSA operates on a safe harbour principle where companies are not responsible for content on their platforms if that content is uploaded by users. Chatbots complicate this framework because conversations between a person and a machine cannot easily be classified as user-generated content.

João Pedro Quintais, a lecturer in law at the University of Amsterdam, described ChatGPT as a hybrid technology combining search engine functions, platform characteristics and features closer to a publisher of its own content. None of the existing DSA categories fully captures this range of uses.

The search engine classification may limit the Commission's ability to control how ChatGPT manages risks related to adolescent mental health, election integrity and illicit content. Experts acknowledge that European regulators have not yet found satisfactory answers to the questions raised by generative AI.

The AI Act runs parallel to DSA obligations

While the DSA focuses on how ChatGPT is deployed to users, the European Union's AI Act regulates the underlying models. Since August 2025, companies developing general-purpose AI models capable of performing diverse tasks must evaluate and mitigate risks these models present. The Commission began enforcing this legislation in late August 2026.

OpenAI, along with Anthropic and Google's Gemini, develops what the AI Act terms general-purpose AI models with potential systemic risks. These companies now face obligations under both the DSA and the AI Act, though the two regulations address different aspects of AI deployment and development.

The dual regulatory framework reflects the EU's attempt to govern AI from multiple angles. Yet the ChatGPT classification demonstrates that existing legislation drafted before the generative AI boom requires interpretation that leaves gaps. The Commission's decision will be watched closely by other technology companies developing similar tools.

People mentioned

Organisations

European Commission · OpenAI · European Parliament