EU AI Act: Which Systems Are Covered?

Understanding the EU AI Act’s Definition of an AI System

The European Union has recently released clarification regarding the definition of an AI system as outlined in its new AI Act. This regulatory framework, which employs a risk-based approach to governing artificial intelligence applications, was initially enacted last summer.

The first compliance requirements – specifically concerning prohibited applications – became enforceable last weekend. Consequently, accurately assessing whether a given software solution is subject to the Act’s regulations is now a crucial task for AI developers.

Potential Penalties for Non-Compliance

Failure to adhere to the AI Act’s stipulations carries significant financial risk. Organizations found in violation could face penalties reaching up to 7% of their total global annual revenue.

The European Commission’s 13-page guidance document is expected to be thoroughly examined by businesses seeking to ensure compliance.

Non-Binding Guidance and Future Updates

It’s important to note that, similar to previous EU guidance on prohibited uses, this advice is not legally binding. The Commission explicitly states that the guidance is intended to be a dynamic resource.

Updates will be issued as needed, taking into account real-world experience, emerging questions, and novel use cases within the rapidly evolving AI landscape. Therefore, interpreting the law’s application will likely be an ongoing process.

The Complexity of Defining an AI System

The EU acknowledges the inherent difficulty in creating definitive classifications. “No automatic determination or exhaustive lists of systems that either fall within or outside the definition of an AI system are possible,” the guidance emphasizes.

This suggests that a nuanced, case-by-case assessment will be required when determining the Act’s applicability. Companies should prepare for a complex and evolving regulatory environment.