Abstract

The political agreement concerning the Artificial Intelligence Act (AIA) has just been reached and although the final wording of the provisions at the time of writing this article was yet unknown, we tried to indicate how this Act and the Digital Services Act (DSA) would interplay. In particular, we have concluded that AI services are not intermediary services, as well as that the limitations of liability under the DSA may be potentiality impacted if the provider of the intermediary services uses the AI systems that require human oversight or a risk-mitigation tool (e.g. content filtering). Non-compliance with the transparency requirement to label the AI generated content may lead to the qualification of such content as illegal under the DSA, while the mere fact that deepfakes are marked in accordance with the AIA does not exclude that they will still be regarded illegal content under the DSA. There are also additional obligations related to the use of AI in recommendation systems by VLOPs and in risk assessments under the AIA as compared with the DSA. Moreover, as AI is changing the way online search engines work, the new search engines may fall outside the application of the DSA.