A European Parliament committee voted to strengthen its legislative proposal as it heads toward a final vote in June.
The rapid evolution of artificial intelligence, or AI, has captivated millions with its ability to compose music, write essays, and perfectly mimic the human voice.
But it has also raised fears about its ability to produce convincing fake news and to manipulate public opinion.
That’s why the European Union is racing to draw up rules to regulate it in an ambitious bill called the AI Act.
On Thursday, the European Parliament’s leading committees green-lighted the bill, paving the way for a plenary vote in mid-June.
So, what’s this AI Act all about and what are the implications for European citizens and tech giants?
Four levels of risk
The bill will classify AI systems according to four levels of risk: from minimal to unacceptable.
An unacceptable level means a particular use of AI is banned; for example, the social credit system in China where local governments rank “good” and “bad” citizens.
Other unacceptable uses include real-time biometric identification in public spaces, where AI scans your face and then automatically identifies you.
AI systems used in high-risk categories like employment and education, which would affect the course of a person’s life, will also face tough requirements such as being more transparent and using accurate data.
Violations will draw fines of up to 6 per cent of a company’s annual global revenue.
What are the main issues EU lawmakers face?
One of the most important issues is how to define what AI is.
“On the one side, they need to regulate certain harms and risks, and on the other side, they must be careful not to stifle innovation,” said Johann Laux, an AI regulation expert told Euronews.
“If you define AI too narrowly, you risk missing out on regulating certain harms. If you define it too broadly, then you run the risk that you are overly inclusive and you will stifle innovation”.
And because technology is developing rapidly, another concern that’s frequently raised is whether the AI Act will still be effective in a couple of years.
“A lot of lawyers out there work with laws and codes that are sometimes even hundreds of years old,” explained Dragos Tudorache, a Romanian MEP co-sponsoring the bill.
“And the end, the secret to that is that you have to try always when writing up law to try to capture the essence of the type of economic or social relationship or phenomenon that you are regulating to the point where that would stand the test of time in five, in ten and 15 and 20 years.
“That’s what we’ve tried to do, that irrespective of how the technology would change, that core system of values and obligations will remain unchanged”.
He told Euronews that parts of the bill that could evolve over time have been put in the text which can be adapted with comitology, the process of implementing or modifying EU laws through committees chaired by the European Commission.
Final approval is expected by the end of the year, or early 2024 at the latest, followed by a grace period for companies and organisations to adapt, often around two years.
For more on this story, watch our report from The Cube in the media player above.