The Digital Services Act faces a (crucial) test from ChatGPT

The Commission’s decision to designate ChatGPT as a ‘very large online search engine’ within the meaning of the EU Digital Services Regulation (Digital Services Act – DSA) marks an important milestone in the European strategy for the governance of artificial intelligence. It means that a new hybrid service – which differs significantly from other online platforms and designated search engines – will be subject to the enhanced regime set out in the DSA. This raises a number of questions regarding how the DSA’s rules will apply to this service, as well as how the European Commission will develop new expertise and methodologies for this purpose.

Whilst, on the one hand, this move imposes on OpenAI obligations to assess and mitigate any systemic risks associated with both the dissemination of unlawful content and that of harmful content (such as disinformation), on the other hand, the decision does not appear to be without its legal complications.

The Digital Services Act limits its scope to intermediary services. This is a diverse category, encompassing entities and business models with characteristics that can vary considerably. Search engines are classified as intermediary services that enable users to formulate queries in order to search, in principle, across all websites – including queries based on keywords – and which return results in any format in which information relating to the requested content can be found.

This definition highlights the distinction between online search engines and traditional storage services, which host information uploaded by service users at their request, as is the case with social media platforms. Furthermore, the concept is agnostic with regard to the output produced in response to the user’s request: whilst search engines normally return hyperlinks to third-party content found on the web, there is nothing to prevent the result from taking another form, such as a summary of that content.

At this juncture, however, a potential inconsistency can be discerned with regard to the way chatbots such as ChatGPT operate. The intermediary services paradigm posits that the service provider remains essentially neutral with respect to the content in question, which also implies the absence of editorial responsibility. Search engines, including those already designated under the DSA, operate in the same way. The content is generally produced by third parties, rather than created by the service providers themselves, as would appear to be the case with most of the output from generative artificial intelligence systems. The crux of the problem thus shifts to a delicate fine line, marking the boundary between the simple compilation of third-party content and a more purely editorial activity.