European Union: Reached Council and Parliament agreement on AI Act including service access restrictions for AI systems posing "unacceptable risk"

On 9 December 2023, the Parliament and the Council of the European Union adopted a provisional agreement on the proposal on harmonised rules on artificial intelligence (AI Act). The compromise agreement clarifies the definition of an AI system by aligning it with the OECD's proposed approach, seeking to differentiate AI from simpler software systems. The AI Act specifically exempts systems exclusively utilised for military or defence purposes and those employed solely for research, innovation, or non-professional personal use. The Act includes a ban on AI systems posing "unacceptable risk". The ban is applicable to cognitive behavioural manipulation, the untargeted scrapping of facial images from the internet or CCTV footage, emotion recognition in the workplace and educational institutions, social scoring, biometric categorisation to infer sensitive data, such as sexual orientation or religious beliefs, and some cases of predictive policing for individuals. In terms of enforcement, the AI Act introduces fines for violations, calculated as a percentage of the offending company's global annual turnover. The fines are tiered, EUR 35 million or 7% for prohibited AI applications, EUR 15 million or 3% for violations of the Act's obligations, and EUR 7.5 million or 1.5% for supplying incorrect information. Following the provisional agreement, the next steps involve finalising the document and submitting it to the European Union Parliament and Council for adoption.