EU Lawmakers Vote to Ban Non-Consensual AI Deepfakes in Landslide Decision

The European Parliament on Thursday moved to clamp down on the darker side of AI, voting overwhelmingly to ban software that creates non-consensual sexual deepfakes. 

The decision, passed with 569 votes in favor and only 45 against, is a direct response to a wave of “nudifier” tools that have recently targeted women and children across the continent. The new ban specifically targets AI systems designed “to create or manipulate images that are sexually explicit or intimate and resemble an identifiable real person” without their permission.

While the rules are strict, they won’t affect AI tools that have “effective safety measures” in place to stop users from generating such content.

The move follows a string of controversies, including a high-profile investigation into Elon Musk’s chatbot, Grok, after it was used to generate realistic, explicit images of real people.

While the Parliament is getting tough on deepfakes, it’s giving companies more time to comply with the broader AI Act. Lawmakers agreed to push back several major deadlines to ensure that technical standards and guidance are actually ready before the rules kick in.

Here is how the new timeline looks:

• Nov. 2, 2026: The deadline for companies to start watermarking AI-generated audio, video, and text.
• Dec. 2, 2027: The start date for “high-risk” AI systems, such as those used in biometrics, law enforcement, and education.
• Aug. 2, 2028: The deadline for AI integrated into products already covered by EU safety laws, like medical devices and toys.

The delay is seen as a pragmatic move. By setting fixed dates, MEPs hope to provide “legal certainty” for businesses that were previously staring down an August 2026 deadline.

Support for growing businesses

The update also includes a “digital omnibus” of changes designed to help European startups scale up. MEPs backed measures that allow companies to process personal data specifically to detect and fix bias in their AI, provided they use strict safeguards.

To help companies that are no longer “small” but aren’t yet “giants,” the Parliament extended SME support measures to “small mid-cap enterprises” (SMCs). Additionally, to avoid drowning companies in paperwork, the Parliament argued that the AI Act’s requirements should be less stringent for products already heavily regulated by other EU laws.

What happens now?

The vote isn’t the final word. The European Parliament must now head into negotiations with the European Council, which represents the 27 EU member states, to hammer out the final text of the law. Since both sides appear to be on the same page regarding the deepfake ban and the timeline delays, final adoption is expected to proceed relatively quickly.

For more on how AI-generated deepfakes sparked political scrutiny, read our coverage of lawmakers probing Musk’s Grok.

The post EU Lawmakers Vote to Ban Non-Consensual AI Deepfakes in Landslide Decision appeared first on eWEEK.