Members of the European Parliament (MEPs) have voted in favour of a new set of rules to govern AI in Europe, which if approved, would be the first of their kind in the world.

Under the proposed regulations, obligations for AI developers and users will be based on the level of risk the technology generates, with systems deemed as having an "unacceptable" impact on safety being strictly prohibited.

Scenarios in which AI would be banned include technology that exploits people's vulnerabilities, uses subliminal or "purposefully manipulative techniques", or carries out social scoring, where people are classified based on their social behaviour, personal characteristics or socio-economic state.

Meanwhile, also joining the "high-risk list" are AI systems that influence voters in political campaigns and in recommender systems used by social media platforms.

“We are on the verge of putting in place landmark legislation that must resist the challenge of time," said Italian MEP and co-rapporteur Brando Benifei, following the vote. "It is crucial to build citizens’ trust in the development of AI, to set the European way for dealing with the extraordinary changes that are already happening, as well as to steer the political debate on AI at the global level."

A ban on discriminatory AI

MEPs also want to introduce a ban on intrusive and discriminatory uses of AI, including real-time remote biometric identification systems in publicly accessible spaces, post-remote biometric identification systems (with the exception of law enforcement for the prosecution of serious crimes), and predictive policing systems based on profiling location or past criminal behaviour.

The European Parliament added that members are pushing for the prohibition all forms of biometric categorisation systems using sensitive characteristics, including gender, race, and ethnicity, as well as "indiscriminate" scraping of biometric data from social media or CCTV footage to create facial recognition databases.


ChatGPT and generative AI

Under the new rules, developers of generative foundation models like ChatGPT would need to disclose when content is generated by AI and design the model to prevent it from generating illegal content or publishing summaries of copyrighted data used for training.

They would also need to "assess and mitigate risks, comply with design, information and environmental requirements and register in the EU database".

Before further negotiations on the final form of the new law can begin, the draft mandate must be endorsed by the whole Parliament in a vote expected to take place in June.

Share Story: