The European Parliament has given its approval to the Artificial Intelligence Act, marking a significant milestone in global AI regulation. The Act received overwhelming support, with 523 votes in favor, 46 against, and 49 abstentions. Co-rapporteur Brando Benifei emphasized the Act’s role in reducing risks, promoting opportunities, combating discrimination, and ensuring transparency in AI development.
This pioneering legislation prohibits high-risk AI applications like emotion recognition in workplaces and schools, social scoring, and the creation of facial recognition databases from CCTV footage. Violations of these regulations could result in hefty fines, up to 35 million euros or 7% of annual turnover for companies. Additionally, the Act imposes transparency requirements on AI companies, necessitating the disclosure of copyrighted content used for training models and compliance with data protection laws.
Furthermore, the Act introduces stringent compliance measures for advanced general-purpose AI systems, including model evaluations, risk assessments, and security incident reporting. While the European Council’s final endorsement is awaited, observers anticipate a staged enforcement of the Act’s provisions over the next two years. Once in effect, the Act’s prohibitions on certain AI applications are set to be implemented within six months, followed by the enforcement of rules affecting general-purpose AI within a year, and additional requirements for high-risk AI systems in critical infrastructure within two years.
