> ## 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: Limits on Crime Prediction
- URL: https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-limits-on-crime-prediction/
- Published: 2026-10-03T01:40:24.000Z
- Updated: 2026-10-04T06:47:10.000Z
- Description: When does AI crime prediction become prohibited? Discover the EU AI Act’s limits on individual risk assessments based on profiling, and the role of verifiable evidence and meaningful human assessment.
- Author: Lydia
- Tags: EU AI Act, Artificial Intelligence (AI), Human Rights / Fundamental Rights, Unacceptable Risk / Prohibited AI Practices, AI Governance

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 **individual crime prediction**, 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 These guidelines superseded the February 2025 draft and provide the detailed examples discussed below.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/).

---

## Limits On Individual Crime Prediction

Article [5(1)(d) ](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com)prohibits one of AI’s most controversial potential uses: predicting whether a particular person will commit a criminal offense.

[Recital 42](https://ai-act-law.eu/recital/42/?ref=thedelatorrereview.com) is straightforward:

---

> "Natural persons should never be judged on AI‑predicted behaviour based solely on their profiling, personality traits or characteristics, such as nationality, place of birth, place of residence, number of children, level of debt or type of car, without a reasonable suspicion of that person being involved in a criminal activity based on objective verifiable facts and without human assessment thereof."

---

Several cumulative conditions apply:

- The practice must involve the *"*placing on the market"* *, "putting into service"* *or "use"** of an AI system (see ‘What Conduct Does Article 5 Regulate?’ in [Where Europe Draws the Line](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-where-europe-draws-the-line/)).
- The AI system must **assess or predict the risk that a natural person will commit a criminal offense.**
- The prediction must be **solely based on profiling of a natural person or on assessing their personality traits and characteristics**

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

**The prohibition does not apply where the AI system supports a human assessment of a person’s involvement in criminal activity that is already based on objective and verifiable facts directly linked to that activity.** The sequence matters. The human assessment must already have a factual foundation. The AI system may support that assessment; it cannot generate suspicion solely from a profile and then rely on a person’s review to legitimize the result. Relevant facts must be:

- objective, rather than merely speculative or based on generalized assumptions;
- verifiable, meaning capable of being checked or substantiated; and
- directly linked to the relevant criminal activity.

Commission Guidelines, § 5.2.3, paras. 203 and 206.

> **Example:** Police investigating a possible armed robbery have evidence that two suspects participated in dark-web discussions about purchasing weapons. An AI system combines geospatial information with automated number-plate recognition data to support the investigation. The assessment is not based solely on profiling because objective and verifiable facts already link the suspects to the planned crime.

Other situations in which AI supports a human assessment already grounded in objective and verifiable facts include:

- AI classifies behavior in a crowd as potentially preparatory to a crime, but a trained person meaningfully reviews the classification and the observable conduct before action is taken.
- A judge uses an AI risk-assessment tool during a pretrial detention hearing but bases the decision on legally valid grounds and objective facts concerning the risk of reoffending, absconding or obstructing the investigation.
- A probation officer uses AI to assess the likelihood that a person will violate release conditions, considering prior conduct, compliance with existing conditions, psychological assessments and recommendations from community services.
- Customs authorities use AI to identify shipments requiring inspection based on the nature and value of the goods, supply-chain information, known trafficking routes and prior involvement in import irregularities.

Commission Guidelines, § 5.3.2, para. 214.

> **Practice Tip:** Ask what objective evidence existed before the AI system was used. If the system generated the initial suspicion from a person’s profile or characteristics, later human review may not bring the practice within the exclusion.

**Place-based predictive policing**: Article 5(1)(d) concerns predictions about natural persons. It generally does not prohibit location-based, geospatial or place-based predictions that assess where crime is likely to occur without predicting that a particular individual will offend.

> **Examples:** An AI system identifies city areas with elevated risks of burglary or knife crime to help determine patrol levels; customs authorities identify locations where narcotics or illicit goods are likely to be found based on known trafficking routes; or acoustic sensors identify and locate possible gunshots in real time.

Commission Guidelines, § 5.3.1, para. 212.

However, if a system uses the risk assigned to a neighborhood as part of profiling an individual—for example, predicting that someone will commit a crime because they live in a high-crime area—the system becomes person-based and may fall within [Article 5(1)(d).](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com)

Commission Guidelines, § 5.3.1, para. 213.

> **Practice Tip:** Place-based policing can become individualized crime prediction when location risk is attributed to a particular person. A high-risk neighborhood is not evidence that one of its residents is likely to offend.

**Article 5(1)(d) is not limited to public authorities.** The prohibition may apply where a private entity is legally entrusted with public powers involving the prevention, detection, investigation or prosecution of crime; acts on behalf of a law-enforcement authority; is asked to conduct an individual crime-risk assessment for law enforcement; or has a specific legal obligation to assess the risk that individuals will commit particular offenses.

- **Anti-money-laundering assessments**: Banks and other regulated entities may be legally required to screen customers for money laundering or terrorist financing. An AI system used for that purpose should rely on the objective and verifiable information specified by applicable anti-money-laundering law and remain subject to human assessment and verification. Compliance with those requirements may place the use outside [Article 5(1)(d).](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com)

> **Example:** A private analytics company is asked by law enforcement to combine national registers, banking transactions, communications and geospatial data to predict which individuals are likely to commit human-trafficking offenses. If the assessment is based solely on profiling or personal characteristics, it may be prohibited.

Commission Guidelines, § 5.2.4, paras. 207–208.

> **Practice Tip:** The prohibition applies to private actors as well as law enforcement, but the system’s purpose remains critical. Detecting suspicious transactions to protect a business is not necessarily individual crime prediction; assessing whether a customer is likely to commit a specified criminal offense may be.

**Predictions concerning legal entities**: Article 5(1)(d) applies to predictions concerning natural persons. It generally does not prohibit AI systems that assess whether a company or other legal entity is likely to commit an offense.

> **Example:** A tax or customs authority uses AI to analyze corporate transactions, declarations and customs records to assess whether a company is likely to commit tax or customs fraud.

Commission Guidelines, § 5.3.3, paras. 215–216.

**Administrative offenses**: The prohibition applies to criminal offenses, not merely administrative violations. An AI system used in an administrative investigation to predict minor traffic violations or irregularities in tax, procurement or public expenditure ordinarily falls outside Article 5(1)(d). Commission Guidelines, § 5.3.4, paras. 217–218.

### 4.1\. Assessing or Predicting The Risk of Future Crimes

Individual crime prediction—or *“crime forecasting”*—generally involves using AI and analytical methods to examine large volumes of historical information and identify patterns associated with criminal activity. The system then generates a risk score predicting whether a particular person is likely to commit or recommit an offense. The information analyzed may include police records, socio-economic information, past behavior and other personal data. These systems may be used to:

- plan law-enforcement operations;
- monitor people classified as high-risk;
- select individuals for inspections or investigation;
- assess the risk of offending or reoffending;
- inform pretrial detention or release decisions; or
- support reintegration planning after a criminal sentence.

Although predictive systems may help authorities allocate scarce resources and take a more proactive approach, relying on historical data can reproduce or reinforce existing biases. A system may also overlook important individual circumstances that are absent from its data or algorithm.

Commission Guidelines, § 5.2.1, paras. 189–190.

**The prediction must concern future criminal conduct.** [Article 5(1)(d)](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com) is concerned with forward-looking assessments of criminal conduct. The assessment may arise at different stages, including crime prevention, detection, investigation, prosecution, sentencing, detention, release and reintegration. Commission Guidelines, § 5.2.1, paras. 191–192.

### 4.2\. **Solely based on profiling of a natural person or on assessing their personality traits and characteristics**

**Profiling and group profiles**: [Article 3(52) AI Act](https://ai-act-law.eu/article/3/?ref=thedelatorrereview.com) defines the concept of “profiling” by reference to [Article 4(4) GDPR](https://gdpr-info.eu/art-4-gdpr/?ref=thedelatorrereview.com). Profiling involves automated processing of personal data to evaluate or predict personal aspects of a natural person, including their work performance; economic situation; health; preferences or interests; reliability or behavior; location; or movements. For [Article 5(1)(d)](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com), the profiling must be used to assess or predict the person’s risk of committing a criminal offense.

- The prohibition can also apply to group profiling. A system may construct a profile of a category of offenders—such as terrorists or gang members—from historical information about crimes committed by other people. If the system later applies that group profile to a particular person to predict that person’s likelihood of committing a similar offense, it is profiling the individual.

Commission Guidelines, § 5.2.2(a), paras. 193–196.

**Personality traits and characteristics:** The prohibition also applies where an individual crime prediction is based solely on personality traits or personal characteristics, even if the processing does not satisfy the technical definition of profiling. The category is broad. [Recital 42](https://ai-act-law.eu/recital/42/?ref=thedelatorrereview.com) gives the following examples:

- nationality;
- place of birth;
- place of residence;
- number of children;
- level of debt; and
- type of car.

Commission Guidelines, § 5.2.2(b), paras. 197–198.

> **Example:** A law-enforcement authority uses AI to predict a person’s likelihood of committing terrorism based solely on age, nationality, address, type of car and marital status. The system treats personal characteristics as predictors of future criminal conduct without objective evidence linking that individual to criminal activity. 

> **Example: The prohibition may also apply when a system assigns risk based on another person’s profile.** A police department uses AI to assess whether children or adolescents are likely to commit future violent or property offenses. A child’s score increases because a sibling, friend or parent has been classified as high-risk. The resulting score leads to police registration, increased monitoring, inspections or referral to youth services.

Commission Guidelines, § 5.2.2(c), para. 202.

**“Solely”**: [Article 5(1)(d)](https://ai-act-law.eu/article/5/?ref=thedelatorrereview.com) prohibits individual crime predictions only when they are based solely on profiling or personality traits and characteristics. That limitation does not mean an organization can avoid the prohibition by adding nominal or irrelevant information to the assessment. Other elements must be real, substantial and meaningful. 

Commission Guidelines, § 5.2.2(c), paras. 199–202.

> **Example:** A tax authority uses AI to examine taxpayers and predict criminal tax offenses using dual nationality, place of birth, number of children and opaque inferred characteristics. Without reasonable suspicion or objective and verifiable facts linking particular taxpayers to criminal activity, the system will normally fall within Article 5(1)(d). It may also constitute prohibited social scoring if it uses information from unrelated social contexts to impose unfavorable treatment.

Commission Guidelines, § 5.2.2(c), para. 202.

> **Practice Tip:** Adding a human reviewer or a few additional data points does not automatically avoid the prohibition. The nonprofiling evidence must be real, substantial and meaningful—not a pretext for a prediction that remains effectively based on personal characteristics.

**NOTE: Falling outside Article 5 therefore does not make the system unregulated.** An individual crime-risk system that falls within the exclusion for human assessments based on objective facts may still be classified as high-risk under Annex III, point 6(d), when intended for use by or on behalf of law-enforcement authorities. Other relevant laws include the GDPR; the Law Enforcement Directive, including its prohibition against profiling that produces discrimination; national criminal-procedure and police laws; rules governing detention, sentencing, probation and release; and Directive (EU) 2016/343 on the presumption of innocence. Commission Guidelines, § 5.4, paras. 219–221.

> **Practice Takeaway:** Article 5(1)(d) draws the line between predicting criminality from who a person appears to be and using AI to help evaluate evidence of what the person has actually done. A system that predicts offending solely from profiles, associations or personal characteristics may be prohibited. A system that meaningfully supports a trained human assessing objective and verifiable facts may instead be high-risk—but permitted subject to the AI Act and other applicable safeguards.

---

## **Other Resources and Recommended Reading**

- [**EU AI Act — Individual Crime Prediction**](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-5?ref=thedelatorrereview.com) — Article 5(1)(d) prohibits specified predictions based solely on profiling or personal characteristics. [Recital 42](https://ai-act-service-desk.ec.europa.eu/en/ai-act/recital-42?ref=thedelatorrereview.com) explains the distinction from objective, verifiable facts.
- [**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 individual risk assessment, profiling, the human-assessment exception and distinctions involving location-based analysis. See § 5, paragraphs 183–221.
- [**EU AI Act — High-Risk Law-Enforcement Systems**](https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng?ref=thedelatorrereview.com) — Article 6 and Annex III, point 6(d), address certain law-enforcement risk assessments that fall outside the prohibition. See recital 59.
- [**Law Enforcement Directive**](https://eur-lex.europa.eu/eli/dir/2016/680/oj/eng?ref=thedelatorrereview.com) — Articles 3(4), 10 and 11 address profiling, sensitive data and automated decisions. Recital 38 explains safeguards and discriminatory profiling.
- [**Presumption of Innocence Directive**](https://eur-lex.europa.eu/eli/dir/2016/343/oj/eng?ref=thedelatorrereview.com) — Articles 3 and 6 protect the presumption of innocence and govern the burden of proof; recital 22 explains that burden.
- [**Charter of Fundamental Rights of the European Union**](https://eur-lex.europa.eu/eli/treaty/char%5F2012/oj/eng?ref=thedelatorrereview.com) — Articles 47–48 protect effective remedies, a fair trial, the presumption of innocence and the rights of the defense.

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

- ***Prohibited AI Practices Under the EU AI Act: Where Europe Draws the Line*** — Explains the scope of Article 5 and the conduct it regulates. [**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: 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: Emotion Recognition at Work and School*** — Examines prohibited emotional inferences from biometric data in workplaces and educational institutions, including the medical and safety exceptions. [**Read the Article.**](https://www.thedelatorrereview.com/prohibited-ai-practices-under-the-eu-ai-act-emotion-recognition-at-work-and-school/)
- ***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-5.png)