Prohibited AI Practices

Article 5 establishes the AI Act's "red lines"—eight categories of AI practices so harmful to fundamental rights, democracy, and human dignity that they are completely banned. There is no compliance pathway for prohibited practices; they simply cannot exist in the EU market.

❗ CRITICAL DEADLINE: Prohibited AI practices must cease by February 2, 2025. This is the first enforcement date under the AI Act. Organizations must complete their prohibition audit immediately.

Understanding the Prohibition Structure

Each prohibition in Article 5 has a specific structure:

1. Conduct: What the AI system does
2. Conditions: Circumstances that trigger the prohibition
3. Harm threshold: Level of harm required (if any)
4. Exceptions: Narrow carve-outs (some prohibitions only)

---

The Eight Prohibited Categories

1. Subliminal and Manipulative Techniques (Article 5(1)(a))

What is prohibited: AI systems that deploy subliminal techniques beyond a person's consciousness, OR purposefully manipulative or deceptive techniques.

Conditions for prohibition:

• The objective OR effect must be to materially distort behaviour
• Must impair the ability to make an informed decision
• Must cause or be reasonably likely to cause significant harm

Example Scenarios:

---

2. Exploitation of Vulnerabilities (Article 5(1)(b))

What is prohibited: AI systems that exploit vulnerabilities of specific persons or groups due to:

• Age (children, elderly)
• Disability
• Specific social or economic situation

Conditions for prohibition:

• Must materially distort the behaviour of a person or group
• Must cause or be reasonably likely to cause significant harm

Example Scenarios:

---

3. Social Scoring (Article 5(1)(c))

What is prohibited: AI systems for evaluating or classifying natural persons or groups over a certain period of time based on social behaviour or personal/personality characteristics, where the social score leads to:

• Detrimental treatment in unrelated contexts, OR
• Treatment that is unjustified or disproportionate to the behaviour

Key nuances:

• The "over a certain period of time" qualifier means that one-off classifications may be treated differently from ongoing surveillance-style scoring
• Applies to both public authorities AND private entities
• The prohibition is about the consequences of scoring, not scoring itself
• Legitimate credit scoring based on financial history remains permitted

Example Scenarios:

---

4. Predictive Policing Based Solely on Profiling (Article 5(1)(d))

What is prohibited: AI systems making risk assessments to predict whether a person will commit a criminal offence, based solely on:

• Profiling, OR
• Personality traits or characteristics

What is NOT prohibited: AI systems supporting human assessment based on objective, verifiable facts directly linked to criminal activity.

Example Scenarios:

---

5. Untargeted Facial Recognition Database Scraping (Article 5(1)(e))

What is prohibited: AI systems that create or expand facial recognition databases through untargeted scraping of facial images from:

• The internet, OR
• CCTV footage

Key elements:

• "Untargeted" is crucial—targeted, lawful collection may be permitted
• Prevents creation of mass surveillance infrastructure
• Applies regardless of whether the database is used

---

6. Emotion Recognition in Workplace and Education (Article 5(1)(f))

What is prohibited: AI systems to infer emotions in:

• Workplace settings, OR
• Educational institutions

Exceptions:

• Medical purposes (detecting pain, fatigue for safety)
• Safety purposes (driver drowsiness detection)

Example Scenarios:

---

7. Biometric Categorisation of Sensitive Characteristics (Article 5(1)(g))

What is prohibited: Biometric categorisation systems that individually categorise persons by inferring:

• Race
• Political opinions
• Trade union membership
• Religious or philosophical beliefs
• Sex life or sexual orientation

Exception: Labelling or filtering of lawfully acquired biometric datasets, or categorisation in law enforcement context.

---

8. Real-Time Remote Biometric Identification in Public Spaces (Article 5(1)(h))

What is prohibited: Real-time remote biometric identification (RBI) in publicly accessible spaces for law enforcement purposes.

Narrow Exceptions (requiring strict conditions):

1. Targeted search for victims of abduction, trafficking, or sexual exploitation
2. Prevention of specific, substantial, imminent threat to life or terrorist attack
3. Locating/identifying suspects of serious criminal offences (Annex II list)

Conditions for exceptions (Article 5(2)-(3)):

• Must be strictly necessary and proportionate
• Deployed only to confirm the identity of the specifically targeted individual
• A completed fundamental rights impact assessment is required (per Article 27)
• Must be registered in the EU database (per Article 49), with an urgency exception allowing post-hoc registration
• Prior authorisation from a judicial authority or independent administrative authority whose decision is binding (or within 24 hours post-hoc in urgent cases)
• If authorisation is rejected, use must stop immediately and all data, results, and outputs must be immediately discarded and deleted
• Temporal, geographic, and personal scope limitations
• No decision producing an adverse legal effect may be taken based solely on the output of the system

Notification and reporting requirements (Article 5(4)-(7)):

• Each use must be notified to the relevant market surveillance authority and the national data protection authority
• Member States may decide to fully or partially authorise real-time RBI use and must lay down detailed national rules, notifying the Commission within 30 days (Article 5(5))
• National authorities must submit annual reports to the Commission on real-time RBI use (Article 5(6))

Annex II Serious Offences include: Terrorism, trafficking, murder, rape, armed robbery, participation in criminal organisation, environmental crimes, and others punishable by a custodial sentence with a maximum period of at least four years (i.e., the maximum possible sentence must be at least four years, not that the minimum sentence is four years).

---

Prohibition Self-Audit Checklist

Use this checklist to audit your AI systems:

Penalties: The Maximum Tier

Prohibited practices carry the highest penalty tier under Article 99:

⚠️ No Transition Period: Unlike high-risk AI requirements, there is NO grace period for prohibited practices. All prohibited AI must cease by February 2, 2025—regardless of when deployed.