AI office must open to society, academia, businesses - MEPs

(ANSA) - ROME, JUL 11 - The European AI Office, set up within the European Commission, must involve civil society, the academic world and companies in drafting codes of conduct for the implementation of requirements for general-purpose AI models (GPAI) and for systemic risk models, according to a letter signed by a group of members of the European Parliament, including an MEP who drafted the AI Act, Brando Benifei, a few weeks after the regulation on artificial intelligence came into effect.
    The AI Act dictates obligations for GPAI model providers and gives the European AI Office the role of implementing and supervising the new rules for such models, with the power of requesting documentation, conducting an evaluation of models, investigating reports and asking providers to adopt corrective measures.
    In particular, GPAI models that don't present systemic risks will be subjected to a few limited requirements, for example on transparency, while the others will have to respect stricter rules.
    In the letter, the MEPs stressed the role the Office can have in "influencing the global governance of AI" with the drafting of codes of conduct for GPAI models and expressed concern for the fact that the Commission intends to "initially involve only interested providers of GPAI models with the risk of enabling them to define concrete practices on their own terms".
    The members of the European Parliament said this "unilateral approach" could "compromise the development of a solid and influential code of conduct at a global level" and thus suggested an "inclusive process" involving "different voices from companies, civil society, the academic world and the interested parties".
    "Enabling companies that dominate the market to influence this process in an isolated way risks creating a restricted and short-sighted perspective that goes against our objectives for EU models of AI development", they observed. (ANSA).