BRUSSELS — Just one year after the European Union adopted a landmark plan to cut the risks of artificial intelligence, it’s already preparing to put the brakes on.

A call to pause the rollout of a large chunk of the EU’s 2024 Artificial Intelligence Act has gained traction over the summer amid heavy lobbying and vocal concerns.

The EU’s executive has left the door wide open for such a pause, and — with a final decision to be taken soon — even those who oppose changing the law are now, in private, factoring in some kind of delay.

One option being discussed could see companies breaching rules on the highest-risk uses of AI being given a pass to continue business as usual for up to a year longer than planned.

A decision to change the law would solidify Europe’s surprising pivot from a global frontrunner in regulating AI to a region fearful of being left behind by the U.S. or China in the race to profit from the technology. 

For many in the industry, it would also validate concerns that the EU failed to deliver a coherent and enforceable rulebook last year, which adequately balances risk with business interests.

Former Italian Prime Minister Mario Draghi, who has written a blueprint on bolstering the EU’s economic prowess, last week slammed the law as a “source of uncertainty” in a surprise intervention that added further momentum.

Drawbacks

Next August is a major date in the law’s staged rollout, with rules kicking in for AI systems that can pose “serious risks” to health, safety and people’s fundamental rights.

Among those are everyday AI tools such as those used in HR, education or the judicial system. Those systems will face various risk management and record-keeping obligations before being put on the market.

But companies are awaiting technical standards that they have not yet received. Industry lobby groups and EU countries have stated that these standards should be ready well ahead of the August deadline so they can comply.

In July, some of Europe’s top CEOs called for a two-year pause “to address the uncertainty.” 

The Commission’s thinking on a potential pause has shifted over the last six months. If standards are not ready in time, “we should not rule out postponing some parts of the AI Act,” the EU’s tech chief Henna Virkkunen told the EU’s digital ministers in June. 

Later, she set a deadline of late August to decide if these standards were ready. 

With no final assessment of the standards in sight, last week saw a fresh acceleration. Draghi declared publicly on Tuesday that the high-risk AI rules should be paused “until we better understand the drawbacks.”

The same day, the Commission opened a consultation on an effort to simplify the EU’s tech rulebooks, in which it said that “targeted adjustments” to the EU’s AI Act are on the table as part of that package. Pausing parts of the act would require an adjustment to the law. 

On Friday, POLITICO reported exclusively on an undated draft of the Commission’s plan to boost AI adoption, expected to be presented Oct. 8. The Commission said the standardization bodies “have not met the deadline to deliver the requested standards.”

The Commission is still keeping its options open, saying in the draft it would not let this development “jeopardise” next August’s enforcement date.

It has declined to be drawn on the next steps. “We have a digital omnibus coming up,” European Commission spokesperson Thomas Regnier told reporters last Tuesday at a daily press briefing, referring to the simplification package that should be finalized in November.

Poland — which helmed the presidency of the Council of the EU until June and has consistently been open to delays — meanwhile circulated a paper to fellow EU countries in which it suggested delaying fines for companies that breach the high-risk AI rules by six or twelve months. 

Warsaw expressed concerns about the law’s “tight implementation deadlines” and warned that European startups could be incited to relocate to “less regulated jurisdictions” if no changes are made. Sweden and the Czech Republic have also shown themselves in favor of a pause.

Hammer to drop

With EU countries set to discuss the Polish proposal on Tuesday, lawmakers within the European Parliament are also preparing for the hammer to drop. 

The Parliament’s top lawmakers on artificial intelligence have called a meeting for Oct. 15 to discuss the effort to simplify the EU’s tech rulebooks and how it would touch on the AI Act. 

One of them, Dutch Greens lawmaker Kim van Sparrentak, is still adamantly against a pause. 

“It’s quite a bizarre proposal to start reviewing this all, while the AI Act has not yet been implemented. We don’t know yet how this law works in practice and how complicated it is,” she told reporters in a briefing last week. 

Yet she acknowledged that damage limitation is now the name of the game.  

“One of the things that we’re increasingly discussing with the people that are focused on the implementation of the AI Act is: how do we ensure that the damage remains limited?”