On 2 August 2026, the European Commission’s AI Act enforcement powers with respect to GPAI model providers, including makers of frontier AI models, entered into force. We co-signed an open letter that examines the current risk and compliance landscape, and that urges the Commission to make full use of its new powers.
The letter notes that several systemic risks, as defined in the EU AI Act, and detailed further in the GPAI Code of Practice, are beginning to concretely materialise. Also, independent analysis based on public information shows gaps between what is required in the Code of Practice and current practices of GPAI providers. Together with the other co-signing parties, we called on the EU AI Office’s Safety Unit A3 to concentrate its resources on these materialising systemic risks related to cyber-offence capabilities, biology capabilities, and loss of control scenarios.