The European Parliament gave the green light to a flagship new artificial intelligence (AI) regulation on Wednesday, paving the way for the law's final adoption by EU member states.
The majority of parliamentarians voted in favour of the law in Strasbourg on Wednesday.
The AI Act is "the first regulation in the world that is putting a clear path towards a safe and human centric development of AI," said EU legislator Brando Benifei in a press conference ahead of the vote.
The forthcoming law establishes a certification regime for AI systems intended for "high-risk" uses listed in the legislation, such as law enforcement, healthcare or hiring and firing.
AI systems intended for such purposes will have to be certified as compliant with the legislation before being put on the market. For example, high-risk AI products must meet certain transparency standards and have adequate risk management systems.
They will also have to be designed to allow human oversight, and the data used to train their algorithms must conform to certain quality standards.
Providers of general-purpose AI systems with diverse potential uses will have to demonstrate that their system doesn't pose a significant risk, or else comply with the rules for high-risk systems.
Some uses of AI will be banned outright, such as systems designed to manipulate people subliminally.
Police use of AI will also be restricted. For example, using AI to identify people from live CCTV feeds would only be allowed in specific circumstances, such as a targeted search for a victim of kidnapping.
The legislation still needs the final approval of EU member states before it can become law. Ministers are expected to vote on the legislation later in the spring.
