by Mariarita Cupersito –
The AI Act, the new European law on artificial intelligence, formally comes into effect today, August 1; however, the effectiveness of the new regulation will be based on the level of risk of the type of AI implemented in the company’s processes.
The first regulatory restrictions will only work from February 2, 2025, and the provisions related to commonly used AI systems will be applied from August 2, 2025. The other provisions that are included in the AI Act effective from August 2, 2026.
At the same time, the European Commission is open to businesses, academia and civil society to draft the code of conduct for generic artificial information providers (GPAI) models: in fact a public consultation has been launched that aims have collected ideas from various stakeholders who come together. the draft code of conduct on GPAI models, a core subject of the AI Act.
A group of MEPs, including the rapporteur of the AI Act Brando Benifei, had written to the European Commission in the last few days to ask for more inclusivity in the process of drafting the code of conduct, which will affect the global governance of Artificial intelligence. The request pointed out that the Community Government seemed inclined to include only interested providers of GPAI models in the first instance, with the real risk of “allowing them the practices concretely define itself”.
The European Office for Artificial Intelligence, established within the Commission, has therefore sought to present contributions to a wider audience of interested parties, including civil society organisations, the world academics, independent experts, rights holders and public authorities, business representatives such as suppliers. of general purpose AI modules or downstream providers integrating such modules into their systems.
The consultation will remain open until September 10 and will cover sensitive areas such as copyright regulations, transparency, identification and risk assessment as well as mitigation and internal management. In this regard, the AI Act considers a two-tiered approach for GPAIs, with stricter obligations for models that pose a systemic risk.
