Unadhi Guides

EU AI Act Article 5(1)(d) criminal-offence risk assessment profiling ban check sheet for accountants, tax advisers and SME clients (solely profiling or personality traits)

Updated 3 Nov 2026 · Disclosure: the EURO AI Register mentioned below is an Unadhi product with a free edition. This guide is for information only, not legal advice.

What you get: a finished Article 5(1)(d) criminal-offence risk assessment / profiling ban check sheet for accountants, tax advisers, bookkeepers, agencies and the SME clients they advise (HR screening, shop security, insurance, credit and law-enforcement-support tools). One row per practice: type in eleven short answers and the sheet returns IN SCOPE OF ART 5(1)(d) BAN, EXCEPTION OR OUT OF SCOPE, CHECK BY HAND or STOP, the status versus today against the 2 February 2025 application date, and a to-do list citing the articles. Rebuild it from the formulas below (LibreOffice-tested). Also included: a short checklist and "what to tell the client" templates. Formulas tested in LibreOffice on twelve made-up rows. Distinct from the Article 5 prohibited list, the Article 5(1)(a) manipulative sheet, the Article 5(1)(b) vulnerability sheet, the Article 5(1)(c) social scoring sheet, the Article 5(1)(f) emotion sheet, and the Article 5(1)(g) biometric categorisation sheet. Note: the Official Journal letter for this ban is Article 5(1)(d); Article 5(1)(g) is the separate biometric categorisation ban.

Deliverable: rebuild the Excel/LibreOffice check sheet from the column list and formulas below (LibreOffice-tested on 12 sample rows + empty row). Keep one row per HR, shop-security, insurance, credit or LE-support practice.

Short answer: from 2 February 2025 (AI Act Article 113), the Union prohibits the placing on the market, the putting into service for this specific purpose, or the use of an AI system for making risk assessments of natural persons in order to assess or predict the risk of a natural person committing a criminal offence, based solely on the profiling of a natural person or on assessing their personality traits and characteristics (Article 5(1)(d); recital 42). The prohibition does not apply to AI systems used to support the human assessment of the involvement of a person in a criminal activity, which is already based on objective and verifiable facts directly linked to a criminal activity. Recital 42 states that the ban does not touch risk analytics that are not based on the profiling of individuals, such as assessing the likelihood of financial fraud by undertakings on the basis of suspicious transactions, or customs tools that predict the localisation of narcotics or illicit goods on known trafficking routes. The Digital Omnibus (Regulation (EU) 2026/1744) added intimate-image and CSAM bans as Article 5(1)(ba)/(bb) from 2 December 2026; it did not amend Article 5(1)(d) (our Omnibus check). Fines for Article 5 breaches sit under Article 99(3): up to EUR 35 000 000 or 7% of worldwide annual turnover, whichever is higher (for SMEs, the lower of those two caps — Article 99(6)).

Why this lands on the accountant's, tax adviser's or agency's desk

Partner programme opening soon

Partner programme opening soon: join the free waitlist. Planned terms are 50% of each referred sale (75% once you have 5+ paying customers), single level, paid through Gumroad only from revenue received.

Step 1: the check sheet

Create a sheet named Practices with headers in row 1 and the first practice in row 2. Create a second sheet named Settings and put =TODAY() in B1 (our test file fixes B1 at 3 Nov 2026). Copy the formulas in L2 to N2 down as far as needed. In Excel 2019 or later, LibreOffice or Google Sheets, TEXTJOIN works as written. The formulas below are the deliverable. Binary workbook kept in the build stage; rebuild from these formulas if you need the file.

ColHeader / formulaNotes
AAI practice / system (HR screening, shop security, insurance, credit, LE support, other)One row per practice
BContext (HR / Shop security / Insurance / Credit / LE support / Customs / Other)Your label
CAI system? (Y/N)Art 3(1)
DPlaced, put into service or used in the EU / output used in EU? (Y/N)Art 2
ERisk assessment of a natural person to assess/predict risk of committing a criminal offence? (Y/N/Unsure)Art 5(1)(d); recital 42
FBased solely on profiling OR assessing personality traits/characteristics? (Y/N/Unsure)Art 5(1)(d); recital 42
GWhich basis? (Profiling / Personality traits / Both / Unsure / None)Art 5(1)(d); recital 42
HClaimed as support for human assessment already based on objective and verifiable facts directly linked to a criminal activity? (Y/N)Art 5(1)(d) exception
IClaimed as non-individual risk analytics (undertaking fraud / customs routes — recital 42)? (Y/N)Recital 42 examples
JStill in use / still offered today? (Y/N)Drives STOP vs IN SCOPE
KFirst used or go-live date (blank if planned only)Date
LArt 5(1)(d) result: =IF(A2="","",IF(C2<>"Y","EXCEPTION OR OUT OF SCOPE: NOT AN AI SYSTEM ON THESE ANSWERS",IF(D2<>"Y","EXCEPTION OR OUT OF SCOPE: NOT IN EU SCOPE ON THESE ANSWERS (ART 2)",IF(OR(E2="Unsure",E2=""),"CHECK BY HAND: CRIMINAL-OFFENCE RISK ASSESSMENT UNCLEAR (ART 5(1)(d))",IF(E2<>"Y","EXCEPTION OR OUT OF SCOPE: NO CRIMINAL-OFFENCE RISK ASSESSMENT ON THESE ANSWERS",IF(OR(F2="Unsure",F2=""),"CHECK BY HAND: SOLELY-PROFILING / PERSONALITY BASIS UNCLEAR (ART 5(1)(d); RECITAL 42)",IF(F2<>"Y","EXCEPTION OR OUT OF SCOPE: NOT BASED SOLELY ON PROFILING OR PERSONALITY TRAITS ON THESE ANSWERS",IF(OR(G2="Unsure",G2="",G2="None"),"CHECK BY HAND: PROFILING OR PERSONALITY BASIS CLASS UNCLEAR (ART 5(1)(d); RECITAL 42)",IF(H2="Y","CHECK BY HAND: HUMAN-ASSESSMENT SUPPORT EXCEPTION CLAIM (ART 5(1)(d)) — KEEP THE OBJECTIVE FACTS",IF(I2="Y","CHECK BY HAND: NON-INDIVIDUAL RISK ANALYTICS CLAIM (RECITAL 42) — KEEP WHY IT IS NOT PROFILING OF PERSONS",IF(J2="Y","STOP: IN SCOPE OF ART 5(1)(d) BAN — SWITCH OFF (LIVE SINCE 2 FEB 2025)","IN SCOPE OF ART 5(1)(d) BAN")))))))))))Ban / out of scope / check / STOP
MStatus vs today: =IF(A2="","",IF(LEFT(L2,9)="EXCEPTION","NO ART 5(1)(d) ACTION ON THESE ANSWERS",IF(LEFT(L2,5)="CHECK","CHECK BY HAND",IF(LEFT(L2,4)="STOP",IF(Settings!$B$1<DATE(2025,2,2),"GAP: BAN APPLIES FROM 2 FEB 2025: "&(DATE(2025,2,2)-Settings!$B$1)&" DAYS LEFT",IF(OR(K2="",K2<DATE(2025,2,2)),"OVERDUE: PROHIBITED SINCE 2 FEB 2025 ("&(Settings!$B$1-DATE(2025,2,2))&" DAYS) — STOP USE NOW","OVERDUE: PROHIBITED SINCE FIRST USE AFTER BAN ("&(Settings!$B$1-K2)&" DAYS) — STOP USE NOW")),IF(LEFT(L2,8)="IN SCOPE",IF(Settings!$B$1<DATE(2025,2,2),"DO NOT SWITCH ON: BAN FROM 2 FEB 2025","DO NOT SWITCH ON: BANNED SINCE 2 FEB 2025 (ART 113)"),"CHECK BY HAND")))))Vs Settings!B1 and 2 Feb 2025
NTo-do list: =IF(A2="","",TEXTJOIN("; ",TRUE,IF(C2<>"Y","not an AI system on these answers (Art 3(1)): no Art 5(1)(d) ban for this row; keep the note",""),IF(AND(C2="Y",D2<>"Y"),"re-check whether the system is placed, put into service or used in the EU, or its output is used in the EU (Art 2(1))",""),IF(AND(C2="Y",D2="Y",OR(E2="Unsure",E2="")),"decide whether the system makes risk assessments of natural persons to assess or predict the risk of a natural person committing a criminal offence (Art 5(1)(d); recital 42)",""),IF(AND(C2="Y",D2="Y",E2="N"),"outside Art 5(1)(d) criminal-offence risk-assessment test on these answers; keep why (e.g. credit default, fraud by undertakings, victim-risk only)",""),IF(AND(C2="Y",D2="Y",E2="Y",OR(F2="Unsure",F2="")),"decide whether the assessment is based solely on profiling of a natural person or on assessing personality traits and characteristics (Art 5(1)(d); recital 42)",""),IF(AND(C2="Y",D2="Y",E2="Y",F2="N"),"not based solely on profiling or personality traits on these answers; keep why (e.g. objective facts already linking the person to a criminal activity). Note: Annex III LE risk tools that are not solely profiling may still be high-risk when permitted — not this ban sheet",""),IF(AND(C2="Y",D2="Y",E2="Y",F2="Y",OR(G2="Unsure",G2="",G2="None")),"name the basis: Profiling / Personality traits / Both (Art 5(1)(d); recital 42 — e.g. nationality, place of birth, place of residence, number of children, level of debt or type of car)",""),IF(AND(C2="Y",D2="Y",E2="Y",F2="Y",G2<>"Unsure",G2<>"",G2<>"None",H2="Y"),"write down the objective and verifiable facts already directly linked to a criminal activity that the human assessment rests on (Art 5(1)(d) closing sentence); get a written expert view before treating it as out of the ban",""),IF(AND(C2="Y",D2="Y",E2="Y",F2="Y",G2<>"Unsure",G2<>"",G2<>"None",H2<>"Y",I2="Y"),"write down why the tool is non-individual risk analytics (recital 42 — e.g. undertaking fraud from suspicious transactions, customs routes for narcotics/illicit goods) and not profiling of natural persons; get a written expert view before treating it as out of the ban",""),IF(AND(OR(LEFT(L2,4)="STOP",LEFT(L2,8)="IN SCOPE"),J2="Y"),"switch the practice off now: Art 5 bans have applied since 2 February 2025 (Art 113); non-compliance with Art 5 can attract fines up to EUR 35 000 000 or 7% worldwide turnover (Art 99(3); for SMEs the lower of the two caps, Art 99(6))",""),IF(AND(OR(LEFT(L2,4)="STOP",LEFT(L2,8)="IN SCOPE"),J2<>"Y"),"do not switch this practice on: Art 5(1)(d) has prohibited it since 2 February 2025 (Art 113)",""),IF(OR(LEFT(L2,4)="STOP",LEFT(L2,8)="IN SCOPE"),"Digital Omnibus Reg (EU) 2026/1744 added Art 5(1)(ba)/(bb) intimate-image and CSAM bans from 2 Dec 2026; it did not amend Art 5(1)(d) criminal-offence risk profiling (our Omnibus check). Note: OJ letter is (d); Art 5(1)(g) is the separate biometric categorisation ban",""),"record who decided, why and the date"))As cited in each item

Step 2: what the test rows returned

Tested in LibreOffice Calc on 3 Nov 2026 with Settings!B1 fixed at 3 Nov 2026, on twelve made-up rows (no real business) plus an empty row. All twelve gave the intended result on the first clean run; the empty row stayed blank; no formula errors. Column grades, the human-assessment-support "CHECK BY HAND" treatment and the non-individual risk-analytics "CHECK BY HAND" treatment are flagged below as ours.

ExampleArt 5(1)(d) resultStatus vs today (excerpt)
1 HR vendor scores applicants' future-theft risk from personality quiz and demographics onlySTOP: IN SCOPE OF ART 5(1)(d) BAN — SWITCH OFF (LIVE SINCE 2 FEB 2025)OVERDUE: PROHIBITED SINCE 2 FEB 2025 (639 DAYS) — STOP USE NOW
2 Lender AI predicts loan-default risk (not criminal-offence risk)EXCEPTION OR OUT OF SCOPE: NO CRIMINAL-OFFENCE RISK ASSESSMENT ON THESE ANSWERSNO ART 5(1)(d) ACTION ON THESE ANSWERS
3 Shop-entrance AI predicts shoplifting risk from face and personality traits onlySTOP: IN SCOPE OF ART 5(1)(d) BAN — SWITCH OFF (LIVE SINCE 2 FEB 2025)OVERDUE: PROHIBITED SINCE FIRST USE AFTER BAN (542 DAYS) — STOP USE NOW
4 LE-support tool claimed to assist human assessment already based on CCTV of the same incidentCHECK BY HAND: HUMAN-ASSESSMENT SUPPORT EXCEPTION CLAIM (ART 5(1)(d)) — KEEP THE OBJECTIVE FACTSCHECK BY HAND
5 Paper staff-vetting checklist (no AI)EXCEPTION OR OUT OF SCOPE: NOT AN AI SYSTEM ON THESE ANSWERSNO ART 5(1)(d) ACTION ON THESE ANSWERS
6 US-only workplace crime-risk AI, no EU useEXCEPTION OR OUT OF SCOPE: NOT IN EU SCOPE ON THESE ANSWERS (ART 2)NO ART 5(1)(d) ACTION ON THESE ANSWERS
7 Planned insurance AI scoring individual "fraudster likelihood" from personality (not live)IN SCOPE OF ART 5(1)(d) BANDO NOT SWITCH ON: BANNED SINCE 2 FEB 2025 (ART 113)
8 Vendor "trust score" for staff; unclear if it predicts criminal-offence riskCHECK BY HAND: CRIMINAL-OFFENCE RISK ASSESSMENT UNCLEAR (ART 5(1)(d))CHECK BY HAND
9 Case tool ranks reoffending using prior convictions and open-case facts (not solely profiling)EXCEPTION OR OUT OF SCOPE: NOT BASED SOLELY ON PROFILING OR PERSONALITY TRAITS ON THESE ANSWERSNO ART 5(1)(d) ACTION ON THESE ANSWERS
10 Customs AI on known trafficking routes (recital 42 non-individual claim)CHECK BY HAND: NON-INDIVIDUAL RISK ANALYTICS CLAIM (RECITAL 42) — KEEP WHY IT IS NOT PROFILING OF PERSONSCHECK BY HAND
11 Insurer AI scores each natural person's insurance-fraud likelihood from personality traits onlySTOP: IN SCOPE OF ART 5(1)(d) BAN — SWITCH OFF (LIVE SINCE 2 FEB 2025)OVERDUE: PROHIBITED SINCE FIRST USE AFTER BAN (581 DAYS) — STOP USE NOW
12 Mall security "crime risk" tags; vendor will not say if solely profiling or personalityCHECK BY HAND: PROFILING OR PERSONALITY BASIS CLASS UNCLEAR (ART 5(1)(d); RECITAL 42)CHECK BY HAND

Step 3: checklist

  1. Is an AI system involved (Articles 3(1), 5(1)(d))?
  2. Is it placed, put into service or used in the EU, or is its output used in the EU (Article 2)?
  3. Does it make risk assessments of natural persons to assess or predict the risk of a natural person committing a criminal offence (Article 5(1)(d); recital 42)?
  4. Is that assessment based solely on the profiling of a natural person or on assessing their personality traits and characteristics (Article 5(1)(d); recital 42)?
  5. Which basis applies (profiling / personality traits / both)?
  6. If a human-assessment support exception is claimed, what objective and verifiable facts already directly linked to a criminal activity are kept in writing (Article 5(1)(d) closing sentence)?
  7. If a non-individual risk-analytics claim is made, why is it not profiling of natural persons (recital 42 — e.g. undertaking fraud from suspicious transactions; customs routes)?
  8. If the row is STOP or IN SCOPE, has the practice been switched off? The ban has applied since 2 February 2025 (Article 113).
  9. Have we told the client that Article 5 fines are the top tier (Article 99(3)/(6))?
  10. Who decided, on what date, and when is the next review?

Step 4: what to tell the client (templates)

Client email — practice looks in scope

Hello [name], from 2 February 2025 the EU AI Act bans AI systems used to assess or predict the risk of a natural person committing a criminal offence based solely on profiling or on assessing personality traits and characteristics (Article 5(1)(d)). The attached one-row-per-practice sheet (eleven answers each) returned STOP or IN SCOPE for [practice]. That is not legal advice. Please switch the practice off (or do not switch it on) and review it with a technology lawyer before any further use in HR screening, shop security, insurance or similar tools.

Client email — out of scope on these answers

Hello [name], on the answers you gave, [practice] sits outside Article 5(1)(d) because [reason from the sheet: not AI / not EU / no criminal-offence risk assessment / not solely profiling or personality]. Keep that reason on file. If the purpose or the data basis changes, re-run the sheet. Ordinary credit-default scoring, and non-individual risk analytics such as undertaking fraud from suspicious transactions or customs tools on known routes, are the recital 42 examples of what this ban does not touch.

Decision record (one per practice)

Practice: [name]. Context: [HR/Shop security/Insurance/Credit/LE support/Customs/Other]. AI system: [Y/N]. EU scope: [Y/N]. Criminal-offence risk assessment of a natural person: [Y/N/Unsure]. Solely profiling or personality traits: [Y/N/Unsure]. Basis: [Profiling/Personality traits/Both/Unsure/None]. Human-assessment support claim: [Y/N + objective facts]. Non-individual analytics claim: [Y/N + why]. Still in use: [Y/N]. Go-live: [date]. Art 5(1)(d) result: [from sheet]. Decided by: [name, role] on [date]. Next review: [date].

Related free guides

Keep each AI practice and its evidence in one place

Record each AI system, each HR, shop-security, insurance or LE-support practice, the Article 5(1)(d) answer and the evidence (decision record, switch-off date, human-assessment or non-individual basis) in the EURO AI Register free edition. It runs in your browser with no account and no upload, covers your first 10 AI systems and 10 evidence records, the screening questions, duties by role with dates, and a printable Article 4 AI-literacy record.

When a client has more than 10 systems, the EURO AI Register membership starts at $19 a month and adds unlimited systems, rule updates and email help. Neither tool is legal advice or proof of compliance.

Partner programme opening soon

Partner programme opening soon: join the free waitlist. Planned terms are 50% of each referred sale (75% once you have 5+ paying customers), single level, paid through Gumroad only from revenue received.

Limits and open points

Read for this guide (3 Nov 2026): AI Act (EU) 2024/1689 Article 2 (scope), Article 3(1) (AI system), Article 5(1)(d) full and the surrounding Article 5(1) list (including the distinct Article 5(1)(c) social-scoring and Article 5(1)(g) biometric-categorisation bans), Article 99(3) and 99(6) as they touch Article 5, Article 113 (Chapters I and II from 2 February 2025), recital 42 on presumption of innocence, solely-profiling / personality-trait criminal-risk assessments and the non-individual risk-analytics examples; Regulation (EU) 2026/1744 (Digital Omnibus on AI) Article 5 amendment inserting (ba)/(bb) only — we searched the Omnibus text for Article 5(1)(d) / criminal-offence risk / profiling letter changes and found none that rename or amend the criminal-risk profiling ban (our check). Fresh EUR-Lex HTML/TXT fetch from this host returned empty/202 on 3 Nov 2026; text confirmed against the same-day CELEX 32024R1689 and 32026R1744 extracts already held from the morning Art 5(1)(a)/(b)/(c)/(g) runs (reused text confirmed today).

Not read for this guide: Commission guidelines on prohibited AI practices (not freely readable from this host today — flag), any Commission FAQ on Article 5, national laws that add or interpret criminal-risk profiling bans, national penalty rules under Article 99(1)-(2), GDPR Article 4(4) / LED Article 3(4) profiling case law beyond the AI Act cross-references, Annex III law-enforcement high-risk offending/reoffending tools beyond noting they are a separate regime for permitted uses that are not solely profiling, and case law on what counts as "based solely on" profiling or personality traits.

Our reading, not the law's words: the column list and grades; treating STOP as "in scope and still live"; treating a human-assessment-support or non-individual analytics claim as CHECK BY HAND rather than automatic out-of-scope; requiring a named profiling/personality basis class before a ban result; ordinary credit-default scoring with no criminal-offence prediction as out of scope on the risk-assessment test; citing the OJ letter as Article 5(1)(d) (not (g)); the sample client wording; and the note that Omnibus did not move Article 5(1)(d).

Frequently asked questions

What does Article 5(1)(d) of the AI Act ban?

It prohibits the placing on the market, the putting into service for this specific purpose, or the use of an AI system for making risk assessments of natural persons in order to assess or predict the risk of a natural person committing a criminal offence, based solely on the profiling of a natural person or on assessing their personality traits and characteristics (Regulation (EU) 2024/1689, Article 5(1)(d)). The prohibition does not apply to AI systems used to support the human assessment of the involvement of a person in a criminal activity, which is already based on objective and verifiable facts directly linked to a criminal activity. Recital 42 explains the presumption of innocence and gives examples of non-individual risk analytics that sit outside the ban.

When did the criminal-offence risk profiling ban start to apply?

Chapters I and II of the AI Act, which include Article 5, apply from 2 February 2025 (Article 113). Non-compliance with Article 5 can attract administrative fines of up to EUR 35 000 000 or 7% of worldwide annual turnover, whichever is higher; for SMEs the lower of those two caps applies (Article 99(3) and 99(6)).

Does every crime-related AI or credit-score tool trigger the ban?

No. Article 5(1)(d) needs a risk assessment of a natural person that assesses or predicts the risk of that person committing a criminal offence, based solely on profiling or personality traits. Ordinary credit-default scoring, and risk analytics that are not based on profiling of individuals (for example undertaking fraud from suspicious transactions, or customs tools on known trafficking routes — recital 42), sit outside this ban on the answers given. Tools that support a human assessment already based on objective and verifiable facts linked to a criminal activity are the Article 5(1)(d) exception and are marked CHECK BY HAND on the sheet.

Did the Digital Omnibus change Article 5(1)(d)?

No. Regulation (EU) 2026/1744 amended Article 5 by inserting points (ba) and (bb) on non-consensual intimate material and child sexual abuse material (from 2 December 2026). We searched the Omnibus text: it does not amend Article 5(1)(d) criminal-offence risk profiling.

How is this different from Article 5(1)(c) social scoring and Article 5(1)(g) biometric categorisation?

Article 5(1)(c) bans social scoring that leads to unrelated or disproportionate detrimental treatment. Article 5(1)(d) bans predicting a natural person's risk of committing a criminal offence based solely on profiling or personality traits. Article 5(1)(g) bans biometric categorisation that deduces or infers listed sensitive attributes. One tool can fail one sheet and pass another; run the matching sheet for the practice you see.

Not legal advice. The sheet helps sort AI practices one by one; it does not decide legal questions. Spotted an error? Write to mininglife7@gmail.com and we correct it.

Sources

← All guides

Unadhi Guides are free, independent explainers. We correct errors as soon as we find them: write to mininglife7@gmail.com.