Revising the Future of AI: Insights from the EU AI Act

The EU AI Act and the European Commission Document C(2025) 924 Final: A Clearer Definition of AI

On February 6, 2025, the European Commission released a pivotal document—number C(2025) 924 final—titled “Guidelines on the Definition of an AI System under Regulation (EU) 2024/1689 (AI Act).” This document aims to clarify what constitutes an AI system, enhancing understanding of the scope of the AI Act, which has been designed to address the rapid proliferation of AI technologies across various sectors.

A Brief History: Why the EU AI Act Came Into Force

The EU AI Act emerged in response to the potential risks posed by AI, including privacy breaches, algorithmic bias, and threats to democratic processes. The European Commission proposed an initial framework in April 2021, leading to extensive dialogue among EU Member States, industry experts, and civil society. Following iterative refinements, the final version of the Act was adopted at the end of 2024, with a phased implementation strategy.

Key Elements Introduced on February 2, 2025

As of February 2, 2025, the initial phase of the EU AI Act came into force, specifically targeting prohibited AI use cases. Key elements include:

• Risk Management Systems: Providers must implement comprehensive risk management procedures covering the entire lifecycle of their AI systems.
• Data Governance and Quality: High-risk systems must adhere to strict standards regarding the quality and governance of their datasets.
• Transparency and Documentation: Detailed technical documentation and operational logs are mandatory.
• Human Oversight: Integral human oversight mechanisms are mandated for intervention and correction of adverse outcomes.
• Conformity Assessments: High-risk AI systems must pass rigorous conformity assessments before entering the market.

These measures are supported by national supervisory authorities across EU Member States, coordinated by the European Artificial Intelligence Board.

Introducing the New Guidelines: (Re)Defining AI Systems

The European Commission document C(2025) 924 final provides essential guidance on the definition of an AI system under the AI Act. The primary goal is to clarify the scope of the Act by outlining seven key definitions:

1. Machine-Based System: AI systems operate on both hardware and software components.
2. Autonomy: These systems function with varying levels of autonomy, performing tasks with minimal human intervention.
3. Adaptiveness: Some AI systems can change behavior post-deployment based on new data.
4. AI System Objectives: AI systems operate according to specific objectives that differ from their general purpose.
5. Inference Capacity: AI systems can infer outputs from inputs using techniques such as machine learning.
6. Types of Outputs: AI systems generate diverse outputs affecting both physical and virtual environments.
7. Impact on Environments: The system must affect physical or digital spaces to be considered AI.

Exclusions from the definition include systems that solely enhance mathematical optimization or use basic data processing without learning capabilities.

What’s Next for the Future of the EU AI Act?

The EU AI Act is designed to evolve with technological progress. Upcoming developments include:

• Transparency for Limited-Risk Systems: Upcoming delegated acts will outline operations for AI systems deemed “limited risk.”
• Enhanced Technical Standards: Further guidelines will address sensitive areas like biometric identification and AI in law enforcement.
• Transition and Support Mechanisms: The European Commission will roll out training programs to facilitate a smooth transition.
• International Alignment: The EU will engage in international dialogues to harmonize standards.

Global Implications of the EU AI Act

The EU AI Act serves as a global benchmark for ethical AI development, impacting:

• EU Entities: Provides legal certainty and simplifies compliance, enhancing market confidence.
• UK Companies: Adherence to EU standards is crucial for market access.
• Global Companies: Non-EU companies must comply to avoid disruptions in international supply chains.

Conclusion

The EU AI Act and the guidelines from document C(2025) 924 final represent a significant milestone in AI regulation. The framework not only addresses prohibited AI systems but also emphasizes the importance of ethical AI development. Businesses globally should pay close attention to the evolving regulatory landscape, viewing it as an opportunity to lead in ethical AI practices.