EU Sets Global Precedent with Comprehensive AI Regulation Act
The European Union has again set a global precedent by unanimously passing a landmark bit of comprehensive legislation to govern AI: The AI Act.
According to the official press release, ministers in the European Parliament passed the AI Act on Wednesday with 523 votes for, 46 against, and 49 abstentions.
The Act brings a targeted regulatory approach to some of the most pressing issues and anxieties around AI today.
To begin with, the Act lists a blanket ban on certain applications of AI that may threaten citizens’ rights, including systems of biometric categorisation “based on sensitive characteristics” and the “untargeted scraping of facial images from the internet or CCTV footage to create facial recognition databases.”
The Act also bans social scoring, emotional recognition at work and in school, and any applications that manipulate or exploit human behaviours and vulnerabilities.
Law enforcement is permitted to use real-time biometric systems with prior authorization “in exhaustively listed and narrowly defined situations” as long as their deployment is limited in scope both in time and in geographical place.
AI use in high-risk areas like critical infrastructure, education and vocational training, employment, essential services (like healthcare and banking), law enforcement, migration and border management, justice and democratic processes are subject to stringent standards of transparency, oversight, reporting and assessment.
General purpose AI is subject to similar requirements. Companies have to comply with European copyright laws and and publish the content they used for training. More powerful general purpose models may need to submit further assessments and reporting
Additionally, all citizens have the right to complain about AI systems and get feedback on any system that has guided a decision that affects their rights. Furthermore, content creators have to clearly label any artificial or manipulated images, audio or video, aka “deepfakes”.
The AI Act From Now Onwards
The AI Act is still not officially through. According to yesterday’s announcement, it still needs a once-over from lawyers and linguists before lawmakers submit it to the European Council for formal endorsement.
The law will then officially come into force after its publication in The Official Journal for the European Union. The bloc’s 27 member states will then have 6 months to enforce the bans on prohibited practises, 9 months to enforce codes of practice, and 12 months for rules concerning general-purpose AI systems.
The AI Act is fully applicable 24 months after entry into force. Businesses have to meet all obligations for high-risk systems within three years of the Act’s official publication.
The EU is keen to get a head start on international rivals like the US and China when it comes to exploiting emerging technologies.
Last year, the bloc passed Markets in Crypto Assets (MiCA), a comprehensive legislative framework for governing cryptocurrencies in Europe that comes into force later this year.
