EU AI Act Compromise: High-Risk Rules Delayed, Nudification Apps Banned

Breaking: EU Reaches Landmark AI Act Deal

Brussels, [date] — Negotiators from the European Parliament and the Council have struck a final compromise on the Artificial Intelligence Act, delaying the high-risk compliance deadline to December 2027 and introducing a long-awaited ban on non-consensual intimate imagery, popularly known as 'nudification' apps.

“This agreement strikes a balance between innovation and fundamental rights,” said a European Commission spokesperson. “We are giving businesses time to adapt while ensuring that the most harmful uses of AI are outlawed.”

The deal follows two failed trilogues and comes as a relief for small and medium-sized enterprises, which will face lighter paperwork requirements under the final text.

Background

The AI Act is the world’s first comprehensive legal framework for artificial intelligence, classifying AI systems by risk level. High-risk applications — such as those used in critical infrastructure, law enforcement, and employment — would have faced strict obligations starting in 2025.

However, intense lobbying from tech companies and member states pushed the compliance deadline back by more than two years. The new deadline now aligns with the 2027 sunset of other digital regulations. Smaller firms, defined as those with fewer than 250 employees and under €50 million in annual turnover, will benefit from simplified documentation and exemption from some costly conformity assessments.

“The delay gives SMEs the clarity they needed to invest in compliance without crushing costs,” said a digital policy advisor from the European DIGITAL SME Alliance. But critics argue the concession weakens the Act’s original ambition.

Nudification Apps Outlawed

The ban on non-consensual intimate imagery — including deepfake pornography — marks a significant victory for digital rights groups. The provision explicitly criminalizes apps that “strip” clothing from images without consent, closing a loophole that had enabled the spread of so-called nudification tools.

“This sends a strong signal that the EU will not tolerate AI being weaponized for harassment,” said a spokesperson for Access Now. The ban takes effect immediately upon the law’s enforcement, likely in 2026. MEP Dragos Tudorache, a co-rapporteur, called it “a red line that must not be crossed in the digital age.”

What This Means

For businesses, the extended timeline provides breathing room to upgrade systems and hire compliance officers. But transparency advocates warn that the delay could leave consumers unprotected against biased or dangerous AI for years. High-risk systems will continue to operate without mandatory human oversight or transparency obligations until the end of 2027.

Consumer organizations have already threatened legal challenges if the delay leads to demonstrable harm. Meanwhile, tech industry associations gave a cautious welcome to the lighter rules for SMEs but expressed concern that the high-risk category remains too broad and could stifle innovation.

The European Parliament is expected to vote on the final text in early March, with adoption likely before the summer recess. The law will then be published in the Official Journal, with most provisions taking effect 20 days later.

This story is breaking. More details will follow.