Eu delays ai rules, bans deepfake nudes in landmark vote
The European Parliament has taken a significant step toward regulating artificial intelligence, approving key revisions to the AI Act Thursday, March 26th. While the sweeping legislation faces further scrutiny, the immediate impact is a delayed rollout of certain high-risk AI applications and a firm prohibition on the creation and dissemination of deepfake pornography.
A pragmatic pause on high-risk ai
Facing concerns that the initial timeline was unrealistic, lawmakers voted overwhelmingly – 569 in favor, just 45 opposed – to postpone the implementation of rules governing “high-risk” AI systems. These include those managing biometric data or deployed in sensitive areas like employment and education. The delay, extending until December 2, 2027, is intended to allow companies time to adequately prepare for the new requirements, a move that signals a more cautious and pragmatic approach to AI governance.
But the Parliament didn't merely punt on difficult decisions. A crucial addition to the AI Act tackles the increasingly prevalent issue of AI-generated non-consensual intimate imagery. The new regulation effectively bans the availability of AI tools capable of generating or manipulating images of a sexual nature depicting identifiable individuals without their explicit consent. A loophole allows for tools with “effective safety measures” to prevent such misuse, demonstrating an awareness of the complexities within the AI development landscape.
Easing the burden on smaller firms
Recognizing the potential for overregulation to stifle innovation, the Parliament also introduced measures aimed at providing greater flexibility for smaller businesses. Companies utilizing AI products that have already undergone regulatory scrutiny in other sectors – such as medical devices or toy safety – will face a lighter touch under the new AI Act. The logic is simple: if a product has already been vetted for safety and compliance, subjecting it to further, redundant oversight is counterproductive.
The revised timeline for watermarking AI-generated content—video, image, audio, and text—has also been extended to November 2, 2026. This requirement, designed to clearly identify AI-generated creations, aims to foster transparency and prevent deceptive use. While the path forward remains complex, this latest vote underscores the EU’s commitment to shaping a responsible and ethical future for AI.
The real test, of course, will be whether these measures can keep pace with the relentless advancement of AI technology, a challenge that demands constant vigilance and adaptation. The clock is ticking.