International Convention on AI: The Council of Europe Publishes a New Compromise Text
The draft, the third one released since the beginning of negotiations, will serve as a basis for the next plenary meeting, scheduled from January 23 to 26. The Council of Europe treaty (not to be confused with the AI Act, the European regulation on artificial intelligence), which applies to the 46 member states of the institution and to a number of observer countries such as the United States, Japan, Mexico, and Israel, aims to ensure that the development, design, and application of artificial intelligence are in line with the principles of the Council of Europe.
Negotiators have reached provisional agreement on a number of issues, including aligning the definition of artificial intelligence systems with the one recently adopted by the OECD. Some provisions are still pending and are the subject of various options in the document. Among these are privacy, the research and development phase, risk management, and prohibited practices.
“It has not yet been decided whether this convention is the best place for provisions that promote security, protection, accuracy, data quality, and other technical and engineering aspects,” analyzes Aleksandr Tiulkanov, a former member of the Council of Europe who has gone into private practice, on his LinkedIn account. The current goal of the parties is to reach an agreement next May.