Alessandro del Ninno will act as the sole speaker at the specialist workshop organised by Paradigma on transparency requirements for AI-generated content and the obligations laid down in Article 50 of the AI Act, in light of the European Commission Guidelines, the Code of Practice and the FAQs concerning transparency of AI-generated content.
The workshop will provide a systematic and practice-oriented analysis of the transparency obligations, starting with their scope of application and the allocation of roles among providers of AI systems, deployers and other operators involved in the production and dissemination of content. Particular consideration will be given to the potential cumulative application of the relevant obligations and to the allocation of responsibilities among providers, business users, agencies, platforms and distributors.
Specific attention will be devoted to the obligations applicable to providers under Article 50(2), with particular reference to AI systems that generate or manipulate text, images, audio or video content, the marking of outputs in a machine-readable format and the availability of tools enabling such outputs to be detected. The workshop will also examine the requirements of effectiveness, interoperability, robustness and reliability applicable to technical solutions, as well as the distinction between content generation, substantial modification, standard editing and merely assistive uses of AI.
As regards deployers, the analysis will focus on the obligations laid down in Article 50(4), with particular reference to the clear and distinguishable disclosure of deepfakes and of AI-generated or AI-manipulated text published for the purpose of informing the public on matters of public interest. Particular attention will be paid to the distinction between the technical marking obligation imposed on providers and the visible or audible disclosure obligation applicable to professional users of AI systems.
A dedicated part of the workshop will address the exemptions and special regimes provided for under Article 50, including the exemption applicable to text that has undergone effective human review or editorial control where a natural or legal person holds editorial responsibility for its publication. The session will examine the minimum substantive requirements of human review, fact-checking processes and the limitations of merely formal or grammatical checks, as well as the regime applicable to artistic, creative, satirical and fictional works.
The programme will also address the practical implementation of Article 50 within media organisations and publishing groups, covering articles, summaries, podcasts, videos, images, newsletters and social media content, as well as the organisation of editorial workflows, source verification, approval processes, the allocation of editorial responsibility, AI interventions occurring after editorial approval, and the retention of evidence documenting the checks performed.
The final part of the workshop will focus on the Code of Practice on the transparency of AI-generated content and on the establishment of an effective compliance framework: the implications of adherence to the Code for demonstrating compliance, alternative approaches available to non-signatories, gap analysis, and the preparation of an implementation plan based on the mapping of AI uses, internal procedures, allocation of responsibilities, disclosure models, controls and documentation of the checks carried out.
