The European Union’s member states have finally approved the world’s first comprehensive AI regulation.
Ambassadors from all 27 countries reached a deal on the landmark AI Act today. The final text on the harmonised rules was agreed after lengthy negotiations between representations of the European Council, European Parliament, and European Commission.
France, Germany, and Italy had all expressed late reservations about the plans. The trio had called for limited regulation of foundation models, a general-purpose AI technology that supports a diverse range of applications. The GPT foundation models, for instance, underpin OpenAI’s ChatGPT.
France had been the last holdout on the deal, but finally dropped its objections today after securing additional conditions. That paved the way for the final agreement.
Lawmakers will hope the pact gives the bloc an edge over its international competitors.
The US has been slower to regulate AI, preferring to let the tech sector push the developments. China, meanwhile, put into effect new rules for generative AI in August, but is yet to approve a comprehensive law for the tech. The EU’s sweeping AI Act is therefore the first of its kind to be adopted.
The bloc will now move to the implementation stage, where lobbying will continue.
Some critics remain concerned about potential changes to the law, while others fear the current rules will inhibit innovation.
“We must stay vigilant in this process to prevent dilution of the EU AI Act’s original intent,” said Bruna de Castro e Silva, AI Governance Specialist at Saidot. “It’s vital to make sure that corporate stakeholders are able to understand the reason for the EU’s policies and what they can do to follow them.
“What we need is transparent guidelines, interweaving governance with the development, deployment, and scaling of AI systems, taking the kind of multidisciplinary and cross-sectorial approach it takes to make real change.”