EU heavies disagree with bloc's AI plans, support UK-style 'wait-and-see' proposal

MEP labels leaked policy 'a declaration of war'

Europe's largest states have put Anglo-centric proposals for light-touch regulation of artificial intelligence on the table in negotiations over the EU's proposed AI Act, which intends to impose harsh rules over developers of the emerging technology.

Germany, France and Italy have declared a new, softer position on AI since they and other EU member states entered negotiations with members of the European Parliament (MEPs), and the European Commission, over proposals to impose sweeping AI laws.

That's according to Reuters and news agencies, to whom the trio leaked copies of their policy paper over the weekend.

The "inherent risk" of AI is in its application, not "the technology itself," the big EU countries state in the paper. They echoed the pro-business policy the UK published in June, and which Prime Minister Rishi Sunak pushed at the Global AI Safety Summit in London early this month.

Laws should therefore not regulate AI, as the EU intends, but its applications, German officials said in statements to Reuters, confirming the proposals set out in the leaked position paper.

That means shelving plans to sanction AI developers for wrong-doing until regulators are able to see there really is a need for it.

"Initially, no sanctions should be imposed," said reports about the paper.

Instead, regulators would develop general principles already being agreed in negotiations at the G7 group of advanced industrial nations, and write those into guidance for AI developers to self-regulate.

The states proposed a compromise that would let the EU impose but one legal demand on AI developers: that they produce "model cards", which describe what the developers are doing in a way that allows public guardians to assess its risk of harm, according to the Euractiv news service.

However, the EU is not taking the leaked paper lightly. An anonymous MEP told Euractiv the proposal was "a declaration of war".

The revision would avert a general AI rule that might hinder innovations in foundation models (general purpose AI technologies that can be turned to diverse applications). Regulators would be told to concern themselves only with those specific uses of AI where it might actually be misused, but has been celebrated for its potential to do things such as spot cancerous cells in medical images and make industrial processes more efficient.

Regulators would keep watch on AI developments, to determine what if any legal mandates might be needed in future to govern its use.

That mirrors the light-touch proposals the UK first set out last year, and wrote into a draft policy in March.

UK AI minister Viscount Camrose restated the UK's wait-and-see policy on AI regulation at a conference held by the Financial Times in London last week. The UK policy of not rushing to regulate AI still stands, he said, a month after prime minister Rishi Sunak himself spelled out the UK position ahead of his meeting with AI experts and policy makers in London.

"The UK's answer is not to rush to regulate. This is a point of principle - we believe in innovation. It's a hallmark of the British economy. We will always have a presumption to encourage it, not stifle it. And in any case, how can we write laws that make sense for something we don't yet fully understand?" Sunak said.