Talks began in Brussels on Wednesday as European Union member states and lawmakers look to iron out some key remaining differences and thrash out an agreement on landmark rules to regulate artificial intelligence (AI).

The closed-door talks, which come after months of negotiations prompted by the bursting onto the scene of ChatGPT last year, are ultimately aimed at establishing the basis on which to approve the world's first comprehensive AI law.

"The world is watching us," said Alexandra van Huffelen, the Dutch minister for digitalization, told the Reuters news agency, adding that it was "critical" for the EU to find a compromise by the end of the year, especially with European parliamentary elections in June.

"Citizens, stakeholders, NGOs and the private sector want us to agree on a meaningful piece of legislation regarding AI," she said.

Why does the EU want to regulate AI?

When ChatGPT was launched almost exactly one year ago, in November 2022, users were wowed by its ability to produce complex works such as poems, essays and speeches within seconds.

Advocates of AI insist that its benefits could go much further, assisting humanity by transforming, streamlining and accelerating everything from work processes to health care.

Critics caution against the unknown risks it could pose to society, disrupting media, judicial systems and democratic processes and sowing unprecedented chaos.

EU bodies say they are not opposed to AI; on the contrary, they want to harness it and support the development of European AI models to compete with those developed in the United States or Asia. The 27-member bloc also envisions regulations for the technology, particularly with regard to EU citizens' data protection.

If it enacts such rules, the EU would become the first political bloc in the world to pass legislation governing AI and could set a precedent for similar models elsewhere.

The longer it delays, the more it risks losing that first-mover advantage, but opponents fear over-regulation could stifle European creativity and development in the AI sector.

What are the stumbling blocks?

There are two main hurdles for lawmakers to overcome regarding the extent to which so-called "foundation models" are regulated and the role of remote biometric surveillance.

The EU's three biggest economies, Germany, France and Italy, are calling for basic foundation models, which are designed to perform a variety of tasks, to be excluded from regulation — and be held accountable instead to less binding codes of conduct.

Experts believe this position is motivated by a desire to avoid hindering the development of European AI models such as France's Mistral AI and Germany's Aleph Alpha.

The second stumbling block pertains to remote biometric surveillance — in other words: facial and other forms of identification via data from cameras in public places.

The European Parliament would like a blanket ban on "real time" remote biometric identification systems, but member state governments want exceptions to be made for national security, defense and military purposes.

How long will talks take?

Discussions are expected to be protracted and could run into the early hours of Thursday morning.

According to the Reuters news agency, which spoke to five people directly involved in the talks, the most likely outcome won't nail down the crucial details but could outline a provisional deal on principles.

A final deal would then need to be agreed before legislation could be put in place, which could pave the way towards it becoming law before European parliamentary elections in June.

But without a deal, the AI Act is likely to be shelved due to a lack of time, bringing to a close EU attempts to get ahead of the curve on the issue as a first regulator.

mf/sms (AFP, Reuters)