The European Union (EU) has reached a provisional agreement on groundbreaking regulations governing the use of artificial intelligence (AI). Following nearly 15 hours of intense negotiations, the deal positions the EU as the first major world power to enact comprehensive laws addressing various facets of AI, ranging from biometric surveillance to the regulation of advanced AI systems like ChatGPT.
European Commissioner Thierry Breton hailed the agreement as a pivotal moment, emphasising the EU's role as a global standard setter. The accord mandates that foundation models, including ChatGPT and general-purpose AI systems comply with transparency obligations before entering the market. This involves creating technical documentation, adhering to EU copyright laws and disclosing detailed summaries of the content used for training.
High-impact foundation models, identified with systemic risk, face additional requirements, including model evaluations, risk assessments, adversarial testing, reporting serious incidents to the European Commission, ensuring cybersecurity and disclosing information on energy efficiency.
Governments within the EU will be restricted in their use of real-time biometric surveillance in public spaces, limited to cases involving victims of specific crimes, prevention of imminent threats, and searches for individuals suspected of the most serious offenses. The agreement explicitly prohibits cognitive behavioural manipulation, untargeted scraping of facial images from the internet or CCTV footage, social scoring and the use of biometric categorisation systems to infer personal characteristics.
Consumers gain the right to launch complaints and receive meaningful explanations, with fines for violations ranging from EUR 7.5 million (USD 8.1 million) or 1.5 per cent of turnover to EUR 35 million or 7 per cent of global turnover.
The legislation is expected to formally enter into force early next year, with implementation slated for two years thereafter.