The European Union boasts that it has beaten the US in the race to regulate how Artificial Intelligence can be used.
Debate has raged for months about balancing the tremendous potential of AI with the dangers it might pose if allowed to rapidly develop unhindered.
But the EU now claims to have solved the conundrum, with the use of AI in fields from policing to healthcare set to be governed across the bloc by new legislation. Tough negotiations between governments and parliamentarians resulted in the Artificial Intelligence Act being passed by a vote of 523 to 46 on Wednesday.
"Europe is now a global standard-setter in trustworthy AI," claimed Thierry Breton, the internal market chief for the European Commission, the appointed body which drafts all legislation for the European Parliament to vote on.
The legislation would impose bans on some “unacceptable” AI uses of the technology and put limits on “high-risk” applications — those in the fields of education, work or critical infrastructure management.
The legislation will ban:
Social scoring systems powered by AI
Use of biometric-based tools to determine political opinions, race or sexual orientation
Use of AI to interpret people's emotions in an education environment or workplaces
Use of some types of automated profiling to predict likelihood of committing crimes
Other regulations set down in the EU AI Act will:
Require new disclosure requirements from companies such as OpenAI
For all AI-generated deepfakes are to be clearly labeled as such
The bill is due to be ratified by the European Council — the meeting of EU member-states' heads of government — in April, and come into force at the end of 2024. Some of the more general-purpose AI rules can be expected to be applied in early 2025, with the remainder set aside until a 2026 application deadline.