> ## Content Index
> Fetch the complete content index at: https://www.thedelatorrereview.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Prohibited AI Practices Under the EU AI Act: Emotion Recognition at Work and School
- URL: https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-emotion-recognition-at-work-and-school/
- Published: 2026-10-03T15:05:57.000Z
- Updated: 2026-10-04T15:33:26.000Z
- Description: The EU AI Act bans biometric emotion recognition in workplaces and schools, with narrow medical and safety exceptions. Explore the ban, practical examples and the limits on monitoring how people feel.
- Author: Lydia
- Tags: EU AI Act, Artificial Intelligence (AI), Unacceptable Risk / Prohibited AI Practices, AI Governance, Biometric Data, Emotion Recognition, Human Rights / Fundamental Rights

The EU AI Act is a **risk-based regulatory framework**, but some practices are considered unacceptable and are **prohibited outright**.

This article examines the prohibition on **emotion recognition at work and school**, drawing on the Commission’s final **July 29, 2025** [**Guidelines on prohibited artificial intelligence practices**](https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act?ref=thedelatorrereview.com). These guidelines superseded the February 2025 draft and provide the detailed examples discussed below.

- For an overview of all prohibited practices, see [Prohibited AI Practices Under the EU AI Act: Where Europe Draws the Line](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-where-europe-draws-the-line/).

[Article 5(1)(f)](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com) prohibits placing on the market, putting into service, or using AI systems to identify or infer a natural person’s emotions in workplaces or educational institutions, subject to narrow exceptions for medical or safety reasons.

The prohibition reflects concerns about both the scientific reliability of emotion-recognition technology and the power imbalance inherent in employment and education. As [Recital 44](https://ai-act-law.eu/recital/44/?ref=thedelatorrereview.com) explains, emotional expressions can vary substantially among cultures, situations and individuals. These systems may therefore have limited reliability, specificity and generalizability while intruding upon privacy, human dignity and freedom of thought and producing discriminatory outcomes.

Four elements define the prohibition:

- The practice must involve the *"*placing on the market"* *, "putting into service"* *or "use"** of an AI system (see, [Prohibited AI Practices Under the EU AI Act: Where Europe Draws the Line](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-where-europe-draws-the-line/) sub-title "What Conduct Does Article 5 Regulate?").
- The system must **infer a person’s emotions from biometric data**.
- The system must be **used in a workplace or educational institution**.
- The use must not fall within the **exception for medical or safety reasons**.

The prohibition applies to providers and deployers within their respective responsibilities.

Commission Guidelines, §§ 7–7.2, paras. 239–243.

**Systems outside work and education**: Emotion-recognition systems used in other contexts do not fall within [Article 5(1)(f)](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com).

Examples may include systems used:

- to analyze customers for marketing;
- to personalize commercial messages;
- to infer customer emotions from voice assistants;
- in intelligent billboards;
- for crowd monitoring at public events;
- by medical professionals during patient examinations; or
- in care robots.

These uses may nevertheless constitute high-risk AI, trigger [Article 50](https://ai-act-law.eu/article/50/?ref=thedelatorrereview.com) transparency obligations, or fall within the prohibitions on harmful manipulation or exploitation. The GDPR, consumer-protection law and other legislation also continue to apply.

Commission Guidelines, § 7.4, paras. 265–269.

> **Practice Takeaway:** Article 5(1)(f) does not prohibit observing workplace or classroom conduct, detecting physical fatigue or analyzing the sentiment of written text. It prohibits using biometric data to identify or infer what workers, candidates or students feel or intend. The medical and safety exceptions are narrow and require a concrete need, strict necessity, proportionality, safeguards and purpose limitation.

## 1\. Inferring Emotions From Biometric Data

The AI Act defines an *“emotion recognition system”* as an AI system intended to identify or infer a natural person’s emotions or intentions from biometric data. ([Article 3(39)](https://ai-act-law.eu/article/3/?ref=thedelatorrereview.com)) .

- “Identifying” an emotion generally involves comparing biometric information—such as a facial expression or voice pattern—with emotions previously programmed into the system.
- “Inferring” goes further. The system analyzes available information, identifies patterns or correlations, and deduces an emotion or intention that is not directly observable from the raw data alone.

Commission Guidelines, § 7.2.1(a)–(b), paras. 244–246.

NOTE: Although [Article 5(1)(f)](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com) refers to systems that “infer” emotions, the Commission interprets the prohibition as covering systems that either identify or infer emotions or intentions

### **Emotions and intentions are interpreted broadly**: 

The relevant emotions are not limited to a closed list. Recital 18 gives examples including:

- happiness;
- sadness;
- anger;
- surprise;
- disgust;
- embarrassment;
- excitement;
- shame;
- contempt;
- satisfaction; and
- amusement.

Organizations cannot avoid the prohibition by describing the output as an “attitude” rather than an emotion. A system that concludes that an employee has an angry attitude toward customers is still making an emotional inference.

Commission Guidelines, § 7.2.1(c), paras. 247–248.

> **Practice Tip:** Review what the system actually infers, not the terminology used by the vendor. Labels such as “engagement,” “sentiment,” “attitude,” “motivation” or “social awareness” may still represent emotional inferences.

### **Observable expressions and physical states**: 

- The mere detection of an observable expression, gesture or movement is not emotion recognition unless the system uses it to identify or infer an emotion or intention. The dividing line is the inference.

> **Examples:** Detecting that a person is smiling is not emotion recognition. Counting how often a television presenter smiles is not emotion recognition. But, concluding from the smile that the person is happy is emotion recognition. Inferring from a student’s voice or gestures that the student is furious and about to become violent is emotion recognition.

- The AI Act excludes physical states such as pain and fatigue from the definition of emotion recognition. [Recital 18](https://artificialintelligenceact.eu/recital/18/?ref=thedelatorrereview.com) expressly identifies systems that detect fatigue in professional drivers or pilots to prevent accidents as examples. Such systems therefore do not constitute emotion recognition for purposes of [Article 5(1)(f)](https://artificialintelligenceact.eu/article/5/?ref=thedelatorrereview.com).

Commission Guidelines, § 7.2.1(c), para. 249.

### **The inference must be based on biometric data**: 

[Article 5(1)(f)](https://artificialintelligenceact.eu/article/5/?ref=thedelatorrereview.com) applies only when the system identifies or infers emotions or intentions from biometric data.   
‘Biometric data’ means personal data resulting from specific technical processing relating to the physical, physiological or behavioral characteristics of a natural person, such as facial images or dactyloscopic data (See, [Article 3 (34](https://artificialintelligenceact.eu/article/3/?ref=thedelatorrereview.com))). Biometric data may arise from physical, physiological or behavioral characteristics, including:

- facial expressions;
- voice characteristics;
- gestures and body movements;
- posture or gait;
- keystroke patterns;
- eye movements;
- heartbeats;
- electroencephalography (EEG); or
- electrocardiograms (ECG).

Commission Guidelines, § 7.2.1(d), paras. 250–252.

> **Example:** An AI system analyzes the content of written text to determine whether an article has a positive or negative tone. Because the inference is based on language rather than biometric data, it falls outside Article 5(1)(f).On the other hand, an AI system infers a person’s emotional state from how they type, their facial expression, body posture or movements. Because those characteristics constitute behavioral or physical biometric data, the system may fall within the prohibition.

Commission Guidelines, § 7.2.1(d), para. 251.

> **Practice Tip:** Ask what input produces the emotional inference. Text-based sentiment analysis is generally outside Article 5(1)(f), while an inference based on typing rhythm, voice, facial expressions, posture or eye movements may use biometric data and fall within the prohibition.

## **2\. Use in a workplace or educational institution**.

### Workplace

*“Workplace”* includes any physical or virtual setting in which people perform tasks or responsibilities for an employer or affiliated organization. The concept can include offices, factories and warehouses; shops, stadiums and museums; outdoor, temporary or mobile worksites; vehicles; remote and hybrid work environments; and videoconferencing and other virtual workspaces.

- The prohibition is **not limited to employees**. It may protect contractors, trainees, volunteers, self-employed people and others working within an organization.
- It also extends to **candidates** during recruitment and selection.

Commission Guidelines, § 7.2.2(a), paras. 253–254.

> **Example:** A call center uses webcams and voice analysis to determine whether employees are angry, impatient or unhappy during customer interactions. The system infers emotions from workers’ biometric data and is prohibited. But, if only deployed for personal training purposes, emotion recognition systems are allowed if the results are not shared with HR responsible persons and cannot impact the assessment, promotion etc. of the person trained, provided that the prohibition is not circumvented and the use of the emotion recognition system does not have any impact on the work relationship.

> **Example:** A supermarket’s or bank’s use of cameras to detect suspicious customers, for example to conclude that somebody is about to commit a robbery, is not prohibited under Article 5(1)(f) AI Act, when it is ensured that no employees are being tracked and there are sufficient safeguards.

> **Example:** A call center analyzes customers’ voices to detect anger or impatience and help employees respond appropriately. Because the system evaluates customers rather than workers, the use is not prohibited under Article 5(1)(f).

> **Example:** An employer uses emotion recognition during recruitment or a probationary period to assess candidates or new workers. Those uses are prohibited.

Commission Guidelines, § 7.2.2(a), para. 254.

> **Practice Tip:** A beneficial workplace objective does not remove the prohibition. Emotion recognition used to improve teamwork, reduce conflict, measure customer service or evaluate workplace culture remains prohibited unless the narrow medical or safety exception applies.

**Incidental capture of workers**: A system directed at members of the public may incidentally capture employees who are present in the same environment—for example, security personnel at a stadium or railway station. The system should generally fall outside Article 5(1)(f) when its primary purpose is not to assess employees’ emotions. The deployer must nevertheless implement safeguards to minimize employee monitoring and ensure that incidental inferences do not adversely affect workers.

Commission Guidelines, § 7.4, para. 270.

> **Practice Tip:** A system aimed at customers or the public should be configured to avoid screening employees. Incidental capture must not become a back door for workplace emotion monitoring.

### Educational Institutions

*"Educational institutions"*are also interpreted broadly to include both public and private educational and training institutions. The prohibition is not limited by the student’s age; the educational level; whether instruction occurs in person, online or through blended learning; whether the institution provides academic, vocational or continuing education; or whether the person has already been admitted.

- It therefore applies to admissions candidates as well as enrolled students.
- Institutions that provide recognized instruction or certification—including vocational schools and continuing-training providers—may fall within its scope.
- Emotion recognition used for educational role-play may fall outside the prohibition when the results are used only for training and cannot affect the participant’s evaluation, admission, certification or other educational outcome. Examples might include actors or teachers practicing how their emotional expressions are perceived.

> **Example:** A school requires students to use a language-learning application that analyzes their biometric data to infer frustration, interest or engagement. The use is prohibited. The same application used independently outside an educational institution does not fall within Article 5(1)(f).

> **Example:** A school uses eye-tracking software during an online examination solely to determine where students are looking and whether they may be consulting unauthorized material. That is not emotion recognition. However, if a system infers anxiety, emotional arousal, or students' interest, attention or engagement, the prohibition applies.

> **Example:** An educational institution monitors both students and teachers. The processing of students falls within the education prohibition, while the processing of teachers also falls within the workplace prohibition.

Commission Guidelines, § 7.2.2(b), para. 255.

> **Practice Tip:** The same application may be lawful in one context and prohibited in another. A consumer may independently use an emotion-enabled language application, but an educational institution generally cannot require students to use it.

## 3\. Use for medical and safety reasons is exempted

Article 5(1)(f) permits workplace and educational emotion **recognition for medical reasons**. The Commission interprets this exception narrowly. Therapeutic uses should generally involve CE-marked medical devices. The exception does not extend to general assessments of well-being; motivation; job satisfaction; learning satisfaction; stress; burnout; or depression.

- A system used to monitor general workplace stress or detect employee burnout remains prohibited. Describing employee monitoring as a wellness program does not make it a medical use.

Commission Guidelines, § 7.2.3, paras. 256–257 and 263.

> **Example:** Emotion recognition assists employees or students with autism or improves accessibility for people who are blind or deaf. A properly designed medical use may fall within the exception.

Commission Guidelines, § 7.2.3, para. 263.

> **Practice Tip:** “Wellness” is not synonymous with “medical.” General monitoring of stress, burnout, depression, motivation or satisfaction does not qualify merely because the organization describes it as supporting health or well-being.

The **safety exception** protects life and health. It does not extend to protecting property or preventing fraud or theft. A qualifying use must address a concrete safety need rather than an employer’s or school’s general interest in efficiency, discipline or risk management.

- Systems that detect physical fatigue in drivers or pilots to prevent accidents generally do not require the safety exception because fatigue is a physical state excluded from the definition of emotion recognition.

Commission Guidelines, § 7.2.3, para. 258 and 261

> **Example:** An employer uses AI to detect whether a worker operating dangerous machinery or handling hazardous chemicals has reached a level of stress or lost concentration in a manner that creates a specific danger. That use may fall within the safety exception. The employer may not reuse the resulting information to evaluate productivity, discipline the worker or make promotion decisions.

Commission Guidelines, § 7.2.3, para. 263.

### **Necessity, proportionality as safeguards**: 

Medical and safety uses must remain strictly necessary and proportionate. Employers and educators should deploy these systems only where an explicit need exists. Data collected for the medical or safety purpose must not be repurposed. Relevant limits include:

- the duration of the monitoring;
- the people subject to it;
- the scale of deployment;
- the seriousness and immediacy of the medical or safety need;
- whether less intrusive alternatives are available; and
- whether the information is confined to the qualifying purpose.

Commission Guidelines, § 7.2.3, paras. 259–260.

> **Practice Tip:** A qualifying safety use should be narrow, temporary and purpose-limited. Document the specific danger, consider less intrusive alternatives and prohibit reuse of emotional inferences for performance management, discipline or other employment or educational decisions.

---

### **FINAL NOTE: Member States may provide stronger worker protections.** 

The AI Act permits the EU and Member States to maintain or introduce laws that provide workers with stronger protection against employer use of AI. More protective collective-bargaining agreements are also permitted.

- A Member State could, for example, prohibit workplace emotion recognition even for certain medical uses that the AI Act would otherwise allow.

Commission Guidelines, § 7.3, para. 264.

## **Other Resources and Recommended Reading**

- [**EU AI Act — Emotion Recognition at Work and School**](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-5?ref=thedelatorrereview.com) — Article 5(1)(f) establishes the prohibition and medical or safety exception. [Recital 44](https://ai-act-service-desk.ec.europa.eu/en/ai-act/recital-44?ref=thedelatorrereview.com) explains the concerns in employment and education / Articles 3(34) and 3(39) define biometric data and emotion recognition. [Recital 18](https://ai-act-service-desk.ec.europa.eu/en/ai-act/recital-18?ref=thedelatorrereview.com) distinguishes emotional inference from observable expressions and physical states.
- [**EU AI Act — Other Emotion-Recognition Requirements**](https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng?ref=thedelatorrereview.com) — Article 6 and Annex III, point 1(c), address high-risk classification; Article 50(3) addresses transparency. See recitals 54 and 132.
- [**European Commission — Guidelines on Prohibited AI Practices**](https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act?ref=thedelatorrereview.com) — Explains the relevant inferences, workplace and educational settings, narrow exceptions and other applicable requirements. See § 7, paragraphs 239–270.

[](https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng?ref=thedelatorrereview.com)

- [**General Data Protection Regulation**](https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng?ref=thedelatorrereview.com) — Articles 5, 6, 9, 22 and 35 address relevant processing, decisions and impact assessments. See recitals 51, 71 and 91.

## **Related The De La Torre Review Articles**

- ***Prohibited AI Practices Under the EU AI Act: Where Europe Draws the Line*** — Introduces the prohibited-practices framework and Article 5’s scope. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-where-europe-draws-the-line/)
- ***Prohibited AI Practices Under the EU AI Act: When AI Persuasion Becomes Prohibited Manipulation*** — Explains the elements separating persuasion from prohibited harmful manipulation and deception. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-when-ai-persuasion-becomes-prohibited-manipulation/)
- ***Prohibited AI Practices Under the EU AI Act: When AI Exploits Vulnerability*** — Examines exploitation linked to age, disability and social or economic circumstances, including the requirements of behavioral distortion and significant harm. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-when-ai-exploits-vulnerability/)
- ***Prohibited AI Practices Under the EU AI Act: The Social Scoring Ban*** — Explains when evaluations of behavior or personal characteristics lead to prohibited adverse treatment and how social scoring differs from legitimate assessments. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-the-social-scoring-ban/)
- ***Prohibited AI Practices Under the EU AI Act: Limits on Crime Prediction*** — Examines the prohibition on predicting individual criminal conduct based solely on profiling or personal characteristics and the exception for supporting human assessments grounded in objective, verifiable facts. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-limits-on-crime-prediction/)
- ***Prohibited AI Practices Under the EU AI Act: Facial-Recognition*** — Explains the ban on untargeted scraping of facial images from the internet or CCTV footage to create or expand facial-recognition databases. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-facial-recognition/)
- ***Prohibited AI Practices Under the EU AI Act: Inferring Sensitive Characteristics From Biometrics*** — Explains the prohibition on individual biometric categorization used to infer specified sensitive characteristics, its exclusions and its relationship with data-protection law. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-inferring-sensitive-characteristics-from-biometrics/)
- ***Prohibited AI Practices Under the EU AI Act: Real-Time Remote Biometric Identification*** — Explains the conditions, exceptions and safeguards governing real-time biometric identification in publicly accessible spaces for law-enforcement purposes. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-real-time-remote-biometric-identification/)
- ***Prohibited AI Practices Under the EU AI Act: Non Consensual Intimate Material and Child Sexual Abuse*** — Examines the new Digital Omnibus prohibitions on generating or manipulating non-consensual intimate material or child sexual abuse material, including provider safeguards and deployer responsibilities. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-non-consensual-intimate-material-and-child-sexual-abuse/)

![](https://storage.ghost.io/c/54/ef/54efeb65-4f2f-479e-b4bb-a9ee526feeaa/content/images/2026/10/ChatGPT-Image-Jul-4--2026-at-04_44_14-PM-7.png)