EU AI Act Begins Enforcement Phase for Prohibited Systems

The European Union's Artificial Intelligence Act began its first phase of enforcement on 2 August, two years after its official entry into force. This initial phase focuses on prohibiting specific AI systems and enacting codes of practice. The Act, which was formally adopted on 13 March 2024 and entered into force on 24 April 2024, establishes a tiered implementation schedule.

Prohibitions against unacceptable AI systems took effect on 2 August 2024. These systems are deemed to pose a clear threat to fundamental rights. Additionally, the same date marked the enforcement of provisions governing codes of practice. These codes aim to guide AI developers and deployers in adhering to the Act's principles.

The majority of the Act's obligations will become enforceable on 2 August 2025. This includes rules concerning high-risk AI systems, which are subject to stringent requirements. These systems necessitate conformity assessments, risk management systems, and human oversight. General-purpose AI models, or GPAI models, will also fall under these regulations next year.

Less than one year remains before these extensive provisions take effect. The final phase of enforcement, pertaining to obligations for high-risk AI systems, is scheduled for 2 August 2026. This staggered approach allows for a gradual adaptation by developers and deployers to the comprehensive regulatory framework.

Member states are currently establishing their national competent authorities. These bodies will be responsible for overseeing the Act's implementation and enforcement within their respective jurisdictions. The European Artificial Intelligence Board (EAIB) and the AI Office are also operational. These entities will play crucial roles in harmonising enforcement across the EU and providing guidance.

The Act’s phased implementation aims to provide a transition period. However, the full operationalisation of all regulatory mechanisms remains several months away. Will the allocated time prove sufficient for comprehensive compliance across all regulated entities?