On 2 February, and after months of negotiation, the European Union’s 27 member states approved the guidelines for the world’s first AI guidelines rulebook and regulations.

The groundwork for the AI Act was first proposed in 2021, when EU policymakers moved to regulate AI systems based on their potential to cause harm. 

Technical refinements of the act took time to finalise, after which the final version of the text was presented on the 24th of January. However, some member states had reservations about the proposed regulations at the time due chiefly to insufficient times for analysis.

Among those with issues were heavyweights France, Germany, and Italy. France, in particular, stood as the primary opponent of the political agreement, advocating for a lighter regulatory regime, especially for powerful AI models. Alongside Germany and Italy, France sought to limit regulatory burdens on promising European startups while ensuring robust rules for high-risk AI models.

With the approval of the AI Act by the Committee of Permanent Representatives last Friday, it appears whatever reservations held previously had been lifted. 

In a statement on X, the Belgian presidency of the Council said the Council of Ministers’ Permanent Representatives Committee had “confirmed the final compromise text found on the proposal on harmonised rules on artificial intelligence”. It described the AI Act as “a milestone, marking the first rules for AI in the EU, aiming to make it safe and in respect of EU fundamental rights.”

Commenting on the proposed AI Act, Nader Henein, VP analyst at Gartner, said that: “(The EU) is putting the best effort forward and trying something out that has a chance. In this particular case, there’s 27 Member States with different agendas, who sat down and agreed to a text that is a lot simpler than people have considered.”

“This idea that regulation needs to be amazing out of the gate is insane,” continued Henein. He said the idea that aiming for a legislation that is flawless would get in the way of progress.

The AI Act essentially says that any organisations wanting to use AI need to assess risks, such as knowing where they’re using AI and what they’re using it for. The act has different tiers ranging from low to high risk – and even prohibited systems. 

One of the main points of dissent for the bill, and regulating technology generally, is that it will obstruct innovation. 

“Innovation, by definition, will find a way,” said Henein. “Innovation always finds a way to work within regulation, rather than to just throw up your hands and give up.” 

“Look at heavily regulated industries like banking, finance, insurance, and healthcare, they operate just fine. These highly regulated industries operate with tonnes of innovation that work with extremely rigid guidelines.”

With the European Parliament’s Internal Market and Civil Liberties Committees slated to adopt the AI rulebook on the 13th of February, followed by a plenary vote provisionally scheduled for 10-11 of April, the formal adoption will then be complete with endorsement at the ministerial level.

“This comes into effect a lot sooner than a lot of people think because of the tiered approach,” said Henein.

“The second it becomes law, we start a six month countdown. At the six month point you’re expected to be compliant in regards to prohibited AI use cases, which also incidentally carry the highest fines.”

While most companies probably won’t have prohibited AI use cases, they must ask the question of themselves nonetheless, since the fine could be 7% of global turnover, or €35 million (£29.9m).

Recommended reading

• Report: Business Leaders Want Sector-specific AI Regulation
• Where Will Companies Look to For AI Regulation Guidance?
• EHRC: UK’s Proposed AI Regulations “Do Not Go Far Enough”

“The harder bar to clear is the one after that, which has come six months later,” Henein points out, which are the high-risk AI use cases. 

“So this is coming fast, a lot faster than a lot of people think. But a lot of them are sitting back and saying ‘oh, but we don’t have any of this.’ That’s an insane assumption,” he warns. 

“AI capabilities have been embedded in almost everything you use, and you need to validate that these embedded AI capabilities are not in violation, which is tough.

“Everything uses AI, everything has AI capabilities, and you’re deploying it. That means you are now a deployer and you are responsible for it.”

As the act comes closer and closer to becoming law, the coming weeks present a vital window for strengthening ethical safeguards while providing flexibility for researchers and companies charting new AI territory. 

As the EU prepares for the implementation phase, Member States will aim to ensure a balanced regulatory framework that fosters AI innovation while maintaining transparency and protecting against potential harms. Slovakia and Austria, among others, have raised concerns and called for clarifications on critical terms and regulatory aspects.

The move by the EU is just one step forward in not just regulating, but understanding how AI will function in our modern society. As the technology becomes more and more synonymous with our everyday life, regulatory bodies around the world must decide what role the technology will play in both business and private life.