France, Germany and Italy Reach AI Regulation Agreement
The governments of France, Germany, and Italy have come to an agreement on regulating generative AI and large language models (LLMs). The three countries issued a joint policy paper that they claim will accelerate AI oversight negotiations at the broader European Union level as well, including mandatory self-regulation for generative AI foundation model developers of AI through ethical codes of conduct.
The rules set out how the governments involved want to steer between stifling innovation while opposing untested regulatory norms on the core technologies themselves. The countries contend that the EU’s AI Act should focus on regulating AI use-cases rather than the underlying technology. Under the agreement, makers of foundation models would be required to provide model cards – datasheets explaining an AI model’s capabilities, limitations, and other details important for proper deployment. An AI governance body could potentially help formulate best practices around model cards. Initially, no penalties would be imposed for violations of the self-regulatory code. However, after an evaluation period, a sanctions-based enforcement system could be introduced.
“The inherent risks lie in the application of AI systems rather than in the technology itself,” the joint paper stated. “The model cards shall include the relevant information to understand the functioning of the model, its capabilities and its limits and will be based on best practices within the developers community.”
The agreement comes as governments globally explore balanced regulations to capture AI’s economic upside while curbing risks like bias. It could help align the EU’s AI Act negotiations between the European Commission, Parliament, and Council. Moreover, the consensus comes as nations balance AI’s prosperity potential with concerns over biases and other risks. It underscores Europe’s influence in establishing principled global norms for emerging technologies while supporting innovation. Italy’s inclusion is especially notable after the country outright banned ChatGPT for a while over concerns about privacy and intellectual property rights. The EU has been rapidly rewriting its proposed AI Act in the wake of the generative AI and synthetic media explosion, testing ideas like requiring developers to list of every piece of copyrighted material used to train their models, among other ideas.