The European Commission has published the Code of Practice on Transparency of AI-Generated Content, developed pursuant to Article 50 of the AI Act.
For those of us who use AI in professional practice, there is one point that deserves careful attention:
Article 50(4) of the Regulation exempts the obligation to label texts as “AI-generated” when two conditions are met: (i) the content has been subject to human review or editorial control, and (ii) a natural or legal person assumes editorial responsibility for the publication.
Insofar as the professional reviews, validates and signs the text under their deontological responsibility, the Code does not require externally labelling each document as “AI-generated”. However, some important qualifications must be noted:
- The exemption requires genuine and substantive editorial control, not merely formal. Signing without genuine review is not sufficient.
- The labelling obligation for texts is limited to those published “with the purpose of informing the public on matters of public interest”. Private professional documents (court briefs, contracts, internal opinions) fall outside that scope.
- The exemption under this Code does not eliminate other deontological obligations specific to each profession, which may require transparency towards the client regardless of what the AI Regulation provides.
The Code also recommends that law firms and professionals who regularly use AI establish proportionate internal processes: identification of the person responsible for editorial oversight, documented review policies, and staff training in the critical use of these tools.
This is a regulatory framework still in development. Prudence advises following its evolution closely.