Germany, France and Italy have agreed on a number of basic rules describing how AI applications should be regulated. The countries strive for self-regulation through codes of conduct for AI basic models.
The agreement states, according to a document seen by Reuters, that developers of basic AI models must provide information about the models in model cards. This must explain, among other things, the operation of the model, as well as its possibilities and limitations. In addition, the countries suggest that an AI supervisor should check these model cards. The AI watchdog could also make efforts to develop guidelines.
For the time being, no sanctions will be imposed under the agreement. If violations of the code of conduct are found, a sanction system may still be introduced. The German Ministry of Economic Affairs believes that AI should not be regulated by laws, but only its application. Otherwise, the trend is that it would limit Europe’s opportunities in the field of AI too much.
Earlier this year, the European Parliament approved the introduction of AI legislation. The AI Act would not come into effect until 2026 at the earliest. The agreement may accelerate negotiations at European level.