Further to the Article, Navigating the EU Act: Ensuring AI Literacy in Legal Practices, this article focuses on the EU AI Act (the Act) and prohibited AI systems.

Article 5 of the Act prohibits certain types of AI systems. These are AI systems that pose an unacceptable level of risk to the health, safety and fundamental rights of individuals. This ban is a prohibition on placing on the market, putting into service or using an AI system for specified use cases.

The ban on prohibited AI systems came into effect on 2 February 2025.

Prohibited AI Systems

The current list of banned AI systems are

1. AI systems that deploy subliminal or manipulative techniques to materially distort behaviour, leading to significant harm;
2. AI systems exploiting vulnerabilities of individuals or group due to their specific characteristics, leading to significant harm;
3. AI systems for social scoring by public and private actors, and not only by or on behalf of public authorities as initially envisaged by the Commission’s 2021 proposal;
4. AI systems to predict criminal offences;
5. AI systems to create facial recognition databases through the untargeted scraping of facial images from the internet of CCTV footage;
6. AI systems inferring emotions in workplaces except for medical or safety reasons;
7. AI systems for biometric categorisation that categorise individually natural persons based on their biometric data to deduce or infer race, political opinion, trade union membership, religious or philosophical beliefs, sex life or sexual orientation; and
8. AI systems for real-time remote biometric identification for law enforcement.

The current list of prohibited AI practices is an initial list that may be added to over time. The Commission will assess the need to amend this list once a year. It is possible that more AI practices may be prohibited in the future.

Guidelines

The European Commission has also published detailed guidelines on the following:

• Guidelines on AI Systems definition under the Act
• Guidelines on prohibited AI practices under the Act

These guidelines are non-binding. The guidelines on the AI Systems definition are designed to evolve over time and will be updated by the Commission as necessary. The guidelines on prohibited AI practices are intended to ensure consistent, effective and uniform application of the Act across the EU. They provide insights into the Commission’s interpretation of the AI prohibited practices under the Act.

At present, the Commission has approved these as draft guidelines but have not yet formally adopted them.

Non-Compliance

Non-compliance with the Article 5 prohibitions is punishable with the highest tier of administrative fines under the AI Act. This includes administrative fines of up to €35M or up to 7 % of the total worldwide annual turnover for the preceding financial year, whichever is higher.

Ireland AI Act Exemptions

Recital 40 of the Act grants certain exemptions to Ireland related to certain AI systems in police and judicial cooperation as set out in Article 5 and 26 of the Act. These exemptions stem from Protocol 21 of the Treaty of the European Union and the Treaty on the Functioning of the European Union.

Protocol 21 governs participation in EU measures related to the areas of Freedom, Security and Justice. Ireland decided to retain control over judicial and police co-operation in criminal matters. This permits Ireland to opt in or remain outside of such measures on a case-by- case basis.

While these exemptions do exist, it is important to note that Ireland may decide to implement the Acts’ regulations relating to law enforcement through national legislation, however it is too early to tell how Ireland will deal with these matters. 

Practical steps that legal practices can take

• Map the use of any systems within your practice and assess whether they fall within the scope of an AI System.
• Map the use of existing AI systems within your practice and assess whether they fall within the scope of any of the prohibited AI systems.
• If an AI system used within the practice is identified as a potential prohibited practice, stop providing or using such system or modify it to ensure compliance with the Act.
• Establish and/or update internal policies and procedures in your practice for monitoring all AI systems to prevent non-compliance with the Act.
• Maintain records of the mapping and assessment exercises, as well as the compliance measures in relation to the assessment of prohibited AI practices.

Carrying out the above steps in relation to prohibited AI practices and putting in place an effective AI Literacy system for your practice will help your practice begin to develop an appropriate AI Governance framework for your practice and assist with compliance with the Act.