European Union legislators have agreed on the details of unprecedented legislation to govern artificial intelligence, paving the way for the world’s most stringent rule on the technology’s growth.
Thierry Breton, EU commissioner, announced the agreement in a post on X.
He referred to the deal as historic. In his writing, he claimed that “the EU becomes the very first continent to set clear rules for the use of AI.” “The AI Act is a springboard for EU start-ups and researchers to lead the global AI race—it’s much more than just a rulebook.”
The agreement came after years of debate among member states and lawmakers about how to limit AI while keeping humanity’s interests at the forefront of the legislation.
It followed long conversations that began this week on Wednesday (06 Dec, 2023).
The new regulation defines “high-risk” AI applications as those that pose a considerable risk to health, safety, fundamental rights, the environment, democracy, and the rule of law. The insurance and banking sectors and election- and voting-related systems are mentioned explicitly as areas requiring stringent safety testing before public deployment.
Citizens can file complaints requesting explanations for how an AI system’s decision has affected their rights and lives.
The package demands deepfake generators, huge language models, and other creation engines to designate their work as AI-generated. It purports to safeguard copyright holders from AI ripping off their work.
The new regulation prohibits using biometric technologies that employ “sensitive characteristics” such as race to identify persons and the indiscriminate scraping of faces from web databases or other picture repositories.
Also prohibited are social credit scoring, “systems that manipulate human behaviour to circumvent their free will,” and emotion recognition by employers or educational institutions.
Law enforcement might only employ face recognition technology under “certain safety and national security exemptions,” like stopping a “specific and present terrorist threat.”
For “post-remote” usage of AI biometric tracking, the target must have been convicted or suspected of committing a serious crime, such as terrorism, drug trafficking, sexual exploitation, murder, kidnapping, rape, armed robbery, participation in a criminal organisation, or “environmental crime.”
The bans will go into effect in six months, while the transparency standards will be implemented in a year. In two years, the entire legislative package will go into effect.
The EU aspires to be a global leader in regulating AI, mirroring its contentious efforts to govern the flow of information on the internet with the Digital Services Act earlier this year.
It, too, carries substantial financial penalties; violators might face fines of up to €35 million ($37.7 million), or 7% of global revenues.
Members of the European Parliament debated their position for years before presenting it to EU member states and the European Commission, the EU’s executive body. Before it becomes legislation, all three parties—countries, lawmakers, and the commission—must agree on the final text.
You will find more information on the EU’s latest legislation on AI here.
Microsoft is working with the EU to remove bundleware from its Windows packages.