Summary of Article 5 of the AI Act, Prohibited Practices, Implementation Steps, and Court Cases
Objectives of the Guidelines and Interpretation
The guidelines have been designed to ensure consistent, effective, and uniform application of the AI law throughout the European Union. While providing legal explanations and practical examples useful to stakeholders, they are not binding; authoritative interpretation is the responsibility of the Court of Justice of the EU (CJEU https://curia.europa.eu/jcms/jcms/j_6/it/). The initiative highlights the EU’s commitment to promoting a safe and ethical AI landscape.
Article 5 of the AI Act: Prohibited AI Practices
The AI Act (https://artificialintelligenceact.eu/) identifies numerous categories of prohibited practices, aimed at protecting individuals and society from the risks posed by potentially dangerous AI systems:
Manipulation and Deception
Subliminal Techniques
The use of imperceptible visual, auditory, or tactile messages that, although not consciously perceived, can influence attitudes or behaviors.
Manipulative Techniques
Practices that alter behavior by exploiting psychological and cognitive vulnerabilities, compromising decision-making autonomy.
Deceptive Techniques
Systems that present false or distorted information, undermining the ability to make free and informed decisions.
Material Distortion of Behavior
The use of such techniques is prohibited if their purpose is to significantly alter the choices and behaviors of individuals (Art. 5, paragraph 1, letters a and b).
Exploitation of Vulnerabilities
Systems may not take advantage of the specific characteristics of vulnerable individuals (due to age, disability, or socioeconomic conditions) to induce harmful choices.
Social Scoring and Criminal Risk Prediction
Prohibitions target the use of AI to classify people unjustifiably or to predict crime risk solely based on automated profiling.
Facial Image Scraping and Emotion Recognition
The indiscriminate collection of biometric data without consent and the use of AI to infer emotions in sensitive contexts are prohibited, except for specific exemptions.
Biometric Categorization and Real-Time Remote Identification
It is forbidden to infer sensitive information (such as race, religion, sexual orientation) and to use biometric recognition in public spaces for law enforcement, with regulated exceptions.
Further Details on Prohibitions
The document elaborates on the characteristics of the prohibited techniques:
At the borders of persuasion:
It is highlighted how the regulation distinguishes between manipulation (which subtly exploits vulnerabilities) and lawful forms of persuasion that, if transparent, do not compromise decision-making autonomy.
Examples of lawful applications:
Some advertising techniques, banking systems, and road safety applications are considered admissible, provided they comply with principles of transparency, data protection, and non-discrimination.
AI Act Implementation Timeline
The application of the AI Act will follow a precise sequence:
2 FEBRUARY 2025:
• Entry into force of the prohibitions (Art. 5)
• Compliance obligation for all AI systems, even if already in use
2 AUGUST 2025:
• Designation of national supervisory authorities
• Entry into force of the sanctioning system
• Start of enforcement powers by authorities
2 AUGUST 2026:
• General application of the AI Act
• Activation of regulatory sandboxes for AI
• Start of compliance obligations for high-risk AI systems
2 AUGUST 2027:
• Application of Article 6(1) for high-risk AI systems
• Deadline for compliance of general-purpose AI models placed on the market before 2 August 2025
Court Cases of AI Hallucinations
The use of AI in the legal field has highlighted the phenomenon of “hallucinations,” i.e., the generation of non-existent or false information:
Italian Case – Florence Court (2025):
A defense attorney used case law references generated by ChatGPT without verifying their authenticity. The Court recognized the error but excluded bad faith and did not apply sanctions pursuant to Art. 96 c.p.c.
American Case – Attorney Steven Schwartz (New York, 2023):
An attorney cited non-existent judgments in a proceeding, leading the judge to sanction the law firm with a $5,000 fine for acting in bad faith and making false statements.
Final Considerations
Control and Transparency:
The adoption of AI systems always requires careful human oversight, cross-checking with reliable sources, and full transparency in the use of tools, as recommended by the “Charter of Principles” of the Milan Bar Association.
Regulatory Commitment:
The AI Act and its guidelines, while providing useful guidance, require final interpretation by the CJEU and the adoption of internal frameworks by companies to ensure compliance and the protection of fundamental rights.
Innovation Challenge:
The integration of safe and ethical AI systems is a challenge where technological innovation must be balanced with respect for ethical and regulatory principles, to avoid manipulative or deceptive practices that could cause significant harm.
2 FEBRUARY 2025
• Entry into force of the prohibitions (Art. 5)
• Compliance obligation for all AI systems, even if already in use
2 AUGUST 2025
• Designation of national supervisory authorities
• Entry into force of the sanctioning system
• Start of enforcement powers by authorities
2 AUGUST 2026
• General application of the AI Act
• Member States must make the regulatory sandboxes for AI operational
• Start of compliance obligations for high-risk AI systems
2 AUGUST 2027
• Application of Article 6(1) on high-risk AI systems
• Deadline for compliance of general-purpose AI models placed on the market before 2 August 2025
Final Considerations
The adoption of AI systems always requires careful human oversight, with cross-checking through reliable databases.
The guidelines aim to ensure uniform and effective application of AI law in the EU. While providing clarifications on prohibitions, they are not binding, with the final interpretation reserved for the CJEU. They provide legal explanations and practical examples to facilitate compliance, reflecting the EU’s commitment to safe and ethical AI.
The guidelines, as well as recommendations developed by the reference authorities of each professional category, such as in the legal field the “Charter of Principles” developed by the Milan Bar Association, emphasize the duty of competence, transparency, and critical oversight in the use of AI.
The implementation of the AI Act, outlined in the timeline above, represents a strict regulatory framework to counter risks and abuses, with possible stricter judicial developments in the future.
Since the AI Act is closely intertwined with data protection, we also invite you to read our Privacy Policy
