KI UPDATE – AI Act steht vor Anpassung

AI Practices
 

The list of prohibited AI practices will be expanded to include sexualized deepfakes: Effective as of December 2nd, 2026, AI systems that generate non-consensual sexual or intimate content, as well as systems that generate depictions of child sexual abuse or assist in their creation, will be banned.

High-risk AI systems
 

Entry into force postponed 

The requirements for product-related high-risk AI systems as set out in Annex I will now apply from August 2nd, 2028, instead of August 2nd, 2027, as previously planned; for context-related high-risk AI systems as set out in Annex III, the requirements will apply from December 2nd, 2027 (previously planned for August 2nd, 2026). This is intended to ensure that technical standards and guidelines that are apparently still missing are available in a timely manner.

Harmonization of sector-specific legislation 

Avoiding duplicate regulation: If sector-specific product legislation under Annex I (e.g., for medical devices, toys, elevators, machinery, etc.) includes comparable AI requirements, the Commission may limit the application of the AI Act. In addition, the Commission is authorized to adopt delegated acts, for example under the Machinery Regulation, to harmonize AI-specific safety requirements for machinery. In such a case, the corresponding high-risk requirements would no longer apply.

Sensitive data

The processing of special categories of personal data for the purpose of detecting and correcting biases is permitted only when absolutely necessary

Registration Requirement 

Contrary to the Commission’s proposal, providers must continue to register their AI systems in the EU database—even if they themselves believe that their system is not a high-risk AI system.

Transparency requirements
 

AI-generated content 

For existing AI systems, the deadline for implementing the labeling of AI-generated content has been extended to December 2nd, 2026. For other transparency requirements and new AI systems, the original effective date of August 2nd, 2026, remains unchanged.

AI real-world labs
 

Implementation deadline postponed 

The deadline for establishing national AI real-world labs has been extended to August 2nd, 2027.

Market surveillance
 

Centralized Oversight by the AI Office 

As proposed by the Commission, the AI Office will assume centralized oversight of AI systems based on general-purpose AI models and developed by the same provider. The same applies to AI systems embedded in very large online platforms or online search engines.

SMEs
 

Relief Measures for SMEs 

Existing relief measures for SMEs will be extended to “small mid-caps” (fewer than 750 employees and revenue of up to EUR 150 million or total assets of up to EUR 129 million) in order to create a more innovation-friendly environment for growing companies.