ANACOM launches a public consultation on draft recommendations for implementing article 5 of the Artificial Intelligence Act

On 16 June 2026, ANACOM launched a public consultation on the draft recommendations for implementing article 5 of the Artificial Intelligence Acthttps://www.anacom.pt/render.jsp?contentId=1790431 (AI Act).

This document sets out organisational recommendations and procedures to support the implementation of the obligation set out in article 5 of the AI Act concerning prohibited artificial intelligence (AI) practices, given the potential risks these practices pose to European values and fundamental rights.

To this end, it takes into account the Guidelines on prohibited AI practices issued by the European Commission (EC), which are intended to ensure consistent, effective and uniform application of the AI Act throughout the European Union (EU), as well as legal explanations and practical examples to help public and private bodies understand and comply with the Act's requirements.

From a subjective perspective, the draft recommendations address the issue of which parties are subject to the prohibitions set out in article 5, specifically the "providers placing on the market or putting into service AI systems or placing on the market general-purpose AI models in the Union, irrespective of whether those providers are established or located within the Union or in a third country" and "deployers of AI systems that have their place of establishment or are located within the Union" or "that have their place of establishment or are located in a third country, where the output produced by the AI system is used in the Union".

From an objective perspective, the draft recommendations analyse the prohibitions set out in article 5, which cover the practices of placing an AI system on the market, putting it into service, and using it, with some variations.

The draft recommendations have been drawn up for the national context and cover both public and private bodies, regardless of their size. They should be taken into account, with consideration given to the specific context of each AI system placed on the market, put into service or used (without precluding the adoption of other appropriate internal policies). Specific reference is also made to solutions with agent-based functionalities that increase criticality and risk, and which therefore require careful analysis of their impact on the risk classification of AI systems.

Bearing in mind the Portuguese context regarding the adoption of AI systems, as briefly described in the document, it is recommended that organisations work together in multidisciplinary teams to develop a plan identifying those AI systems that are prohibited and must therefore cease operating, not be launched and/or be withdrawn from the market.

This plan must fully consider the guidelines issued by the EC and the best practices that can be implemented through a sequential, six-step programme, namely:

• the identification of AI systems;
• the mapping, analysis and classification of AI systems according to risk;
• the identification of prohibited AI systems;
• the registration of prohibited AI systems;
• the removal of prohibited AI systems from the market;
• the annual assessment of AI systems.

These recommendations can, and for reasons of efficiency should, be aligned with the Organisational Guidelines for the Implementation of AI Literacy, so that, should an organisation adopt the steps set out in those guidelines, it can subsequently apply them within the framework of these recommendations, and vice versa, thereby avoiding duplication of procedures and tasks.

The document also refers to the temporal application of the AI Act's provisions, noting that those in Chapter II (including article 5) have been applicable since 2 February 2025. The prohibitions set out in this article will apply to all AI systems, regardless of whether they were placed on the market or put into service before or after this date.

The AI Act provides for the monitoring of compliance with the prohibitions rules and stipulates that non-compliance is subject to fines of up to €35,000 or, if the offender is a company, up to 7% of its global annual turnover in the previous financial year, whichever is higher - thus setting a higher maximum fines for breaches of the prohibitions rules set out in article 5 - clearly demonstrating the European legislator's commitment to ensuring that AI systems are developed and used responsibly and safely, in accordance with the EU's fundamental rights. It also highlights the seriousness of offences that constitute a breach of these rules.

Finally, the possible amendments to article 5 of the AI Act resulting from the Digital Omnibus Proposal are noted. However, organisations are reminded that, regardless of whether any new prohibited practices are identified, it would be prudent and useful for them to adopt and follow best practice in this area with immediate effect.

Interested parties may submit their contributions to this draft set of recommendations in writing and in Portuguese by 16 July 2026, to the email address [email protected].

Once the consultation process is complete, all contributions received will be made public. Interested parties should therefore submit a redacted version of any confidential information for publication on this website.

Consult:

• Draft recommendations for implementing article 5 of the Artificial Intelligence Act https://www.anacom.pt/render.jsp?contentId=1850918