EU AI Act Targets U.S. Tech Giants with Unprecedented Regulations

High-risk AI systems deal with autonomous vehicles, medical products, and related high-risk products, which are appraised based on the risks these pose to the health, safety, and fundamental rights of citizens. This will encompass financial services and education, where risk is ingrained in the AI application.

U.S. Big Tech companies in focus The rules, according to Cripps partner Matthew Holman, would carry significant implications for anyone or any company developing, creating, using, or reselling AI within the EU—putting U.S. tech firms squarely in the crosshairs.

"The EU AI is unlike any law anywhere else on earth," Holman said. "It creates for the first time a detailed regulatory regime for AI."

"U.S. tech giants have been following this developing law very closely," said Holman. "There's been a whole lot of funding put into public-facing generative AI systems, which are going to have to now make sure they're following the new law that is, in some places, quite onerous."

The EU Commission will be able to impose fines on companies that violate the AI Act of up to 35 million euros ($38 million) or 7% of their annual global revenues, whichever is higher.

This development in EU law has followed the launch of OpenAI's ChatGPT in November 2022. At that time, lawmakers realized under the current law, there wasn't enough granularity that would apply to the sophistication of the latest generative AI technology and the risks involved regarding the material protected by copyright law.

A Long Way to Implementation

The terms of reference make it clear that the EU would broadly target the generative AI systems, such as general-purpose AI, among other things. This would require EU copyright laws. They should be transparent in terms of the ways in which models are trained, undergo regular testing, and have sufficient cyberprotection.

'But it will be some time before these requirements really bite,' said Dessi Savova, partner at Clifford Chance. The general-purpose systems will not face constraints for 12 months after AI Act becomes effective.

And even then, today's commercially available generative AI systems like OpenAI's ChatGPT, Google's Gemini, and Microsoft's Copilot take the same principles but with a "sunset clause" and give them another 36 months after the implementation date to bring their technology into legal compliance. "We've agreed on the AI Act, and this rule book is about to become a reality," she said in an email to CNBC. "Now, attention must turn to the effective implementation and enforcement of the AI Act."