EU Advances AI Regulation with Extended Deadlines and Stricter Deepfake Controls

The European Union is moving forward with a revised approach to artificial intelligence regulation, incorporating extended deadlines for high-risk AI systems and introducing targeted restrictions on the generation of deepfake content. These measures are part of a broader legislative overhaul aimed at ensuring the responsible development and deployment of AI technologies within the EU.

Revised Implementation Timeline for High-Risk AI

Under the latest amendments to the AI Act, the timeline for regulating high-risk AI systems has been extended. The Council of the European Union has agreed to postpone the enforcement of certain rules by 16 months compared to previous plans. This adjustment addresses ongoing concerns regarding the absence of standardized frameworks and tools necessary for effectively overseeing high-risk AI applications.

High-risk AI systems are defined as those capable of impacting fundamental rights or public safety. These include applications such as biometric remote identification, including facial recognition, as well as AI utilized in critical infrastructure, judicial processes, and the education sector. The legislation stipulates that AI must not be solely responsible for critical functions, such as exam evaluations in educational settings, without human oversight.

According to the updated regulation, the requirements for standalone high-risk AI systems will take effect from December 2, 2027. For high-risk AI integrated into broader systems, compliance will be mandatory starting August 2, 2028. Operators of such systems will be required to register them in an EU-wide database, regardless of their own assessment of risk level, ensuring transparent oversight and accountability.

Deepfake Content Generation Restrictions

The Council is also taking steps to prohibit the use of AI for generating specific types of deepfake content. The proposed restrictions target the artificial creation of images or videos depicting child abuse, intimate situations, or non-consensual sexual acts. The regulation emphasizes that the lack of consent for digital creation is as significant as the absence of consent for the depicted act itself.

These measures are a response to the growing prevalence of inappropriate and harmful AI-generated imagery, particularly following the misuse of image generators such as Grok, developed by xAI and widely shared on social media platforms. Organizations and advocacy groups have long called for bans on face-swapping applications, which can easily superimpose individuals' faces onto existing explicit content, raising serious ethical and privacy concerns.

Enhanced Oversight for General Purpose AI Models

The AI Office, an EU regulatory body, will maintain primary oversight of general purpose AI models. The latest proposal also requires that any exceptions, where national authorities may have jurisdiction, be clearly enumerated by this central office to avoid regulatory ambiguity.

In addition, the Council has asked the European Commission to develop guidelines aimed at reducing administrative burdens for businesses deploying AI systems. These guidelines are expected to extend simplified requirements, previously available only to small and medium-sized enterprises, to certain mid-cap companies, supporting innovation while ensuring compliance.

Next Steps in the Legislative Process

The Council's recommendations and amendments will now be discussed with the European Parliament as part of the ongoing legislative process. The goal is to finalize a regulatory framework that both addresses emerging risks associated with AI and supports the dynamic growth of the sector within the European single market.