The European Commission has designated OpenAI’s ChatGPT as a "very large online search engine" under the Digital Services Act (DSA). That classification subjects ChatGPT to the same transparency and risk-mitigation obligations that apply to search services with more than 45 million monthly active users, such as Google Search or Bing. Failure to comply could expose OpenAI to fines of up to 6 percent of its total global annual turnover.
Narrow scope: conversational features fall largely outside the designation
The Commission’s decision applies specifically to those ChatGPT functions that operate in a search-like manner. In practice, free-form conversational interactions where the system generates its own answers are largely considered outside the strictest DSA obligations.
Christel Schaldemose, a Danish Social Democrat Member of the European Parliament and one of the DSA’s principal authors, warned that ChatGPT is far more than a simple search engine. She argued that some of the most serious harms associated with chatbots — notably emotional dependency among children and manipulative interface designs — may therefore escape the strongest regulatory safeguards.
Real-world risks and context
AI-based chatbots have been the subject of intense debate worldwide after concerns that such systems may have played a role in teenagers’ self-harm in several cases; the widely reported example of Adam Raine in California is often cited. An industry insurance survey found that about 60 percent of adults use chatbots for therapeutic or therapy-like purposes, raising questions about safe usage and mitigation measures.
A concrete regulatory gap highlighted by observers is this: asking ChatGPT for a list of local election candidates would likely fall under the DSA’s remit, while a conversational exchange about who to vote for — a context where misinformation and manipulation are more likely — would probably lie outside the DSA’s search-focused coverage.
Parallel regulation: the AI Act and enforcement steps
Beyond consumer-facing rules, the Commission also seeks to supervise general-purpose AI models such as ChatGPT via the proposed AI Act. Since August 2023, OpenAI and other large model developers have been required to assess and mitigate systemic risks posed by their systems; the Commission began enforcing those duties at the end of August this year, holding hearings with several AI companies about their safety procedures.
Daniel Leufer of Access Now criticized the current risk-assessment focus for prioritizing existential threats — for example, nuclear or biological risks and loss of control — over the protection of fundamental rights and everyday societal harms.
João Pedro Quintais, a law professor at the University of Amsterdam, described ChatGPT as a hybrid technology that functions simultaneously as a search engine, a social platform and a publisher of original content. That hybridity means the search-engine designation could limit the Commission’s supervisory reach in critical areas such as election integrity or the handling of illegal content.
Brando Benifei, the European Parliament’s co-rapporteur on AI legislation, acknowledged that DSA-based oversight is urgently needed to protect users and argued that the DSA and the AI Act can be complementary regulatory frameworks.
Alternatives and implications
The Commission could have chosen to classify ChatGPT as a very large online platform instead, which would have triggered stricter content-moderation duties. However, platform designation also leans on the "safe harbor" principle that can shield companies from liability for user-uploaded content. In the case of chatbots, it remains contentious whether human–machine dialogues should be treated as user-generated content at all.
Conclusion
The Commission’s ruling marks a significant regulatory step by bringing some ChatGPT functions under the DSA’s requirements for very large search engines. Nonetheless, experts and lawmakers caution that many of the risks arising from conversational chatbot interactions — including mental-health harms and potential manipulation of public opinion — remain largely outside the DSA’s present scope, indicating a need for further legal and supervisory measures.
