MEPs have made their proposed amendments to the EU AI Act including tightening rules on the use of biometric facial scanning in public places.
The European Commission has come closer to fully implementing its AI Act, which has been in the works for some years now.
The internal markets committee and the civil liberties committee adopted a draft negotiating mandate AI regulations, with 84 votes in favour, 7 against and 12 abstentions.
If the draft negotiating mandate is endorsed by the whole European parliament, the laws will be passed. The final vote is expected next month.
As far back as 2021, the AI Act was being touted as a way of reigning in what the EU saw as ‘high-risk’ AI activities. The AI Act proposal was made in the same year. Mainly, the EU wants to ensure that citizens’ rights will be protected as the tech advances.
In a series of amendments made to the proposal, MEPs argued that AI systems would work for people and be safe, transparent, non-discriminatory and environmentally friendly.
They also argued for a uniform definition for AI that will be relatively neutral so it can be applied to the tech as it evolves over time.
In the works for a long time
The Commission has already made attempts to work ethical AI values into its law-making. Last September, for example, it released the AI Liability Directive as a way of protecting consumers’ legal rights against companies that sell AI-powered products and services.
The old EU consumer liability laws were outdated and did not cover the possible harmful effects on consumers of AI tech, such as drones.
The Commission has also been called upon to regulate AI in policing where tech such as biometric facial recognition scans are used. Last November, a group of MEPs proposed a complete ban on this type of tool being used by authorities for fear of abuse.
Today, the amendments included a ban on the discriminatory use of AI tech such as remote biometric identification systems in public places.
Biometric categorisation systems that use sensitive characteristics like gender and race will be banned, as will predictive policing AI tech.
The Commission voted that the right to privacy will be upheld with the scraping of biometric data from social media or CCTV footage to create facial recognition databases also banned.
Any application for AI that might be seen as ‘high-risk’ will be carefully regulated. Generative AI models, like ChatGPT, will have to comply with transparency requirements, including disclosing that its content was generated by AI. Companies using generative AI will also have to disclose summaries of copyrighted data used for training.
There will be some exemptions to the limits on AI, especially where innovation is concerned.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.