The EU AI Act: New Requirements Coming into Force on February 2, 2025

The EU’s Artificial Intelligence Act (AI Act) entered into force on August 1, 2024, but most provisions will only become applicable in phases. As of February 2, 2025, the first wave of requirements came into force, introducing critical obligations regarding development, deployment and use of artificial intelligence (AI) for companies operating within the EU.

1. AI Literacy Requirements

One of the most notable requirements that will take effect on February 2, 2025, is the obligation for companies to ensure AI literacy among their staff. Article 4 of the AI Act mandates that all providers and deployers of AI systems must ensure that their employees possess a sufficient level of knowledge and understanding regarding AI technologies. This includes awareness of both the opportunities and risks that AI presents.

In practical terms, this requirement necessitates the implementation of AI governance policies and training programs tailored to the specific needs of the organization. Companies may need to develop internal guidelines, standards, and training courses aimed at equipping staff with the necessary skills to engage with AI technologies responsibly. This obligation applies even to organizations using low-risk AI systems, emphasizing that understanding AI is crucial across all sectors.

2. Prohibited AI Practices

Another major aspect of the AI Act effective from February 2, 2025, is the prohibition of certain AI practices deemed to pose unacceptable risks. Article 5 outlines a list of prohibited uses of AI technologies that violate fundamental rights or ethical standards. Key prohibitions include:

• Subliminal Techniques: AI systems that manipulate individuals’ behavior through subliminal or deceptive techniques are banned. This includes practices like neuromarketing that influence consumer decision-making without conscious awareness.
• Exploitation of Vulnerabilities: The use of AI to exploit vulnerabilities related to age, disability, or socio-economic status is prohibited. For instance, content targeting children in a manipulative manner falls under this ban.
• Social Scoring: The Act forbids systems that evaluate or classify individuals based on their social behavior or characteristics, leading to detrimental treatment—such as governmental social scoring systems.
• Predictive Policing: Conducting risk assessments solely based on profiling or personality traits to predict criminal behavior is not allowed.
• Inferences about Emotions: AI systems that infer emotions in workplaces or educational institutions are banned unless justified for medical or safety reasons.

These prohibitions apply to both providers of such systems and companies utilizing them. Violations may lead to substantial fines, emphasizing the serious nature of compliance.

Enforcement Mechanisms

The enforcement of the AI Act will be executed through a complex web of national regimes combined with EU-level oversight. Each EU member state is tasked with designating competent authorities responsible for enforcing compliance with the AI Act by August 2, 2025. Countries have flexibility in structuring their enforcement frameworks; for example:

• Centralized Approach: Some countries may establish a dedicated agency (like Spain’s AI Supervisory Agency) to oversee compliance across all sectors.
• Decentralized Model: Other nations might employ existing regulators from various sectors (e.g., health or safety authorities) to monitor compliance.

The penalties for noncompliance vary based on the nature of the violation:

1. Engaging in prohibited AI practices can attract fines up to €35 million or 7% of a company’s worldwide annual turnover.
2. Noncompliance regarding high-risk AI obligations may result in fines up to €15 million or 3% of worldwide turnover.
3. Providing incorrect information to authorities can lead to fines up to €7.5 million or 1% of worldwide turnover.

Unlike other EU regulations such as the General Data Protection Regulation (GDPR), the AI Act does not offer a one-stop-shop mechanism for cross-border enforcement; however, it establishes a European Artificial Intelligence Board for coordination among national authorities.

Preparing for Compliance

As companies gear up for the implementation of these new requirements, they should take proactive steps to ensure compliance with the AI Act:

1. Assess Current AI Systems: Businesses should evaluate their existing AI technologies and practices against the prohibitions outlined in the Act. This includes identifying any systems that may exploit vulnerabilities or engage in prohibited practices.
2. Develop AI Literacy Programs: Organizations must create training programs that enhance employees’ understanding of AI technologies and their implications. Tailoring these programs based on staff roles and responsibilities is essential.
3. Establish Governance Policies: Implementing internal governance frameworks to manage AI system usage effectively can help ensure adherence to new regulations.
4. Monitor Regulatory Developments: Companies should stay informed about guidance from national authorities and the European Commission regarding compliance strategies and best practices.

Conclusion

The upcoming implementation of the EU AI Act marks a pivotal moment in the regulation of artificial intelligence within Europe. With significant requirements coming into effect on February 2, 2025, organizations must prepare for enhanced obligations around AI literacy and prohibitions on high-risk uses of technology. As companies navigate this complex regulatory landscape, proactive measures will be essential in ensuring compliance and fostering a culture of responsible AI use.

By adhering to these new standards, businesses not only mitigate legal risks but also contribute positively to an ethical framework that prioritizes safety and fundamental rights in the development and application of artificial intelligence technologies. The path ahead requires diligence, adaptability, and a commitment to understanding the evolving realm of AI regulation in Europe.