The European Parliament has ruled: artificial intelligence applications generating non-consensual sexual content are now persona non grata across the continent. This decision marks the culmination of a targeted revision of the AI Act, Europe’s landmark AI legislation, passed this Tuesday. The goal? To rebalance an already ambitious legislative framework, easing certain burdens for businesses while bolstering fundamental safeguards.
This update, adopted by 423 votes for, 57 against, and 174 abstentions, is part of the “digital omnibus package.” It signals a clear intent: to foster innovation, but not at any cost. Essentially, Europe is fine-tuning its regulatory approach to AI, adapting to the breakneck pace of technological advancements.
Timelines Relax, But Vigilance Remains
One of the key changes involves pushing back several compliance deadlines. Obligations for “high-risk” artificial intelligence systems will now take effect from December 2, 2027, specifically for autonomous systems. For those integrated as safety components and already governed by other sectoral laws, the effective date is August 2, 2028. This move aims to prevent an excessive compliance burden on already regulated products.
General Entry into Force
Application of most provisions of the initial AI Act.
AI Content Marking & “Nudifying” Compliance
Mandatory digital labeling of content generated by AI systems. Companies must comply with the ban on “nudifying” applications.
High-Risk Autonomous AI Systems
Obligations begin for these specific systems. A delay to better adapt.
AI Systems Integrated as Safety Components
Application for systems already covered by other sectoral legislations. The last step in the current timeline.
The deadline for digital marking of AI-generated content has also been extended, now set for December 2, 2026. By this date, all algorithmically produced content must be identifiable via machine-readable labeling. The goal? To boost transparency and enable users to differentiate between authentic and synthetic content, a critical challenge as AI-generated images and texts proliferate across the web.
A Strong Signal Against Abusive Uses
But perhaps the most significant development is the ban on so-called “nudifying” applications. To be clear: this refers to systems capable of creating, without any consent, explicit sexual images, videos, or audio, or depicting intimate parts of an identifiable person. It’s essentially preventing AI from becoming a tool for widespread digital harassment.
Formal Ban on “Nudifying” AIs
Providers will be prohibited from marketing these systems within the European Union unless they implement robust technical safeguards to prevent the creation of child sexual abuse material or non-consensual sexual content. Users leveraging these tools for malicious purposes are also explicitly targeted.
Affected companies have until December 2, 2026, to bring their systems into compliance. This move directly addresses escalating concerns over deepfakes and other image manipulation technologies. Europe is sending a clear, unequivocal message: ethics take precedence, particularly when human dignity is on the line. While GDPR previously established foundational principles for personal data protection, this represents a crucial specification for generative AI capabilities.
Streamlining Responsible Innovation
Beyond the prohibitions, the AI Act introduces significant simplification measures. Certain requirements, deemed redundant for AI-integrated machinery, have been eliminated, provided that existing sectoral safety rules ensure an equivalent level of protection. The definition of a “safety component” is also clarified to prevent the automatic classification of mere assistance or optimization functions as high-risk if they pose no inherent danger.
✅ Key Strengths
Regulatory Clarity: Fewer redundancies for already secure systems, freeing up developers.
SME Support: Specific exemptions for small and medium-sized enterprises facilitate their development in AI.
Bias Correction: Authorization to process personal data to detect and correct biases, under strict conditions.
⚠️ Areas for Concern
Persistent Complexity: Despite simplifications, the framework remains dense and demanding for stakeholders.
Constant Adaptation: The rapid evolution of AI could render some aspects obsolete even before their full application.
These new provisions also permit, under strict conditions, the processing of personal data to detect and correct biases within AI systems — a significant stride towards fairer AI. Additionally, exemptions have been broadened for small and medium-sized enterprises to fuel their growth. This is vital for startups, including promising French tech companies, which innovate but often struggle to navigate a sea of regulations.
As co-rapporteur Arba Kokalari emphasized, this text aims to “facilitate European innovation while upholding the regulation’s core objectives.” Michael McNamara, her counterpart, highlighted the importance of legal certainty and crucial protection against non-consensual imagery. Before officially taking effect, this legislation still awaits formal adoption by the Council. Yet the message is unambiguous: Europe is forging ahead, committed to a regulated, human-centric, and responsible AI future.
Chargement de la galerie…
About Rigaud Mickaël
Passionate about tech and a Linux enthusiast, I decipher AI with a unique and intense vision to make it useful to all, between robots, rock and the geek universe.
