Germany, France and Italy have reached an agreement on how artificial intelligence should be regulated in the future according to a joint paper. The three governments support "mandatory self-regulation through codes of conduct" for so-called foundation models of AI. Developers of foundation models would have to define model cards, which provide information about a machine learning model's functioning, capabilities, limits and are based on best practices. An AI governance body could help develop guidelines and check the application of model cards. Initially, no sanctions should be imposed if violations are identified, but after a period of time, a sanctions system could be established.