A proposal drafted by the European Commission on the regulation of artificial intelligence was leaked ahead of its publication date. The official announcement is expected to be next week.
According to the draft, the European Union plans to avoid the dystopian scenarios from Black Mirror or Westworld by taking a strict approach to AI. In fact, the EU is planning to ban AI systems that are used to manipulate people’s behavior or opinions or exploit their vulnerabilities, and systems used for mass surveillance or social scoring.
As for AI applications classified as high-risk AI, additional conformity requirements will be applied. High-risk AI includes systems that pose a direct safety threat or have a high chance of affecting people’s livelihood.
Systems deemed as high-risk AI include technologies that are currently in daily use across Europe. Companies are widely using algorithms to e.g. scan CVs or job applications or assess credit worthiness.
Companies failing to comply with the new rules could face a fine up to 20 million eur or 4% of their turnover.
The regulation will not apply to systems used by military or authorities safeguarding public security.
The proposal draft has already been criticized by experts for being too open to interpretation. European policy analyst Daniel Leufer expressed concern on its ‘loopholes and vagueness’ in his tweet.
The EU’s approach differs from that of the US and China, where the regulation is much less strict. This could affect the technology industry in Europe by making the EU area less appealing market for tech companies.
It is not known if changes have been made to the proposal after the leak. It will also take several years for the proposal to become binding legislation.