EU AI Act Article 5(1)(ba) non-consensual intimate deepfake ("nudify") ban check sheet from 2 December 2026 for accountants, tax advisers and SME clients
What you get: a finished Article 5(1)(ba) intimate deepfake ban check sheet for accountants, tax advisers, bookkeepers, agencies and the SME clients they advise (photo apps, avatar and image generators, marketing agencies, photo studios, chatbots that make images). One row per tool: type in eleven short answers and the sheet returns IN SCOPE, EXCEPTION OR OUT OF SCOPE, CHECK BY HAND or STOP, the days left to 2 December 2026, and a to-do list citing the articles. Rebuild it from the formulas below (tested in LibreOffice on twelve made-up rows). Also included: a checklist and a "what to tell the client" email.
Deliverable: rebuild the Excel/LibreOffice check sheet from the column list and formulas in Step 1.
Short answer: the Digital Omnibus, Regulation (EU) 2026/1744, inserts a new prohibited practice into Article 5(1) of the AI Act, point (ba): AI systems that generate or manipulate realistic images, videos, audio or similar material of an identifiable person's intimate parts, or of that person engaged in sexually explicit activities, without that person's freely-given, specific, informed, unambiguous and explicit consent. It applies from 2 December 2026 (Article 113(a) as amended). For providers it bites when that output is the intended purpose, or a reasonably foreseeable and reproducible outcome with no reasonable and adequate safeguards (new Article 5(1a)(a)). For deployers it bites only when they use the system for that purpose (Article 5(1a)(b)). Edits that do not increase exposure of intimate parts or change the nature of a depicted sexual activity are not "manipulation" (Article 5(1b)). The same amendment adds point (bb) on child sexual abuse material under Directive 2011/93/EU; that point is not graded by this sheet and any such case goes straight to the police and a lawyer. Fines for Article 5 breaches are the top tier (Article 99(3), lower cap for SMEs under Article 99(6)).
Why this lands on the accountant's or agency's desk
- 23 days left (as of 9 Nov 2026). Clients that sell or embed image, avatar or "photo enhance" features into EU products need a written answer before 2 December 2026.
- Software houses and app studios are usually the provider. Their question is whether their safeguards are documented and tested, not whether the feature exists.
- Agencies, studios and ordinary SMEs are usually deployers. Their exposure is a staff member misusing a tool, so a written use policy matters.
- It sits next to the Article 50(4) deepfake disclosure sheet and the Article 50(2) marking sheet, which share the 2 December 2026 date for older generators. Labelling content never makes a banned output lawful.
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 Tools with headers in row 1 and the first tool in row 2. Create a sheet named Settings and put =TODAY() in B1 (our test file fixes B1 at 9 Nov 2026). Copy L2 to N2 down. TEXTJOIN works in Excel 2019 or later, LibreOffice and Google Sheets.
| Col | Header / formula | Notes |
|---|---|---|
| A | AI tool or practice | One row per tool |
| B | Context (Image app sold to customers / Marketing agency / Photo studio / Chatbot with images / Other) | Your label |
| C | AI system? (Y/N) | Art 3(1) |
| D | Placed, put into service or used in the EU, or output used in the EU? (Y/N) | Art 2 |
| E | Can it produce realistic images, video or audio of an identifiable person's intimate parts or sexually explicit activity? (Y/N/Unsure) | Art 5(1)(ba); Art 5(1b) |
| F | Client's role (Provider / Deployer / Unsure) | Art 3(3)/(4) |
| G | Provider only: is that output the intended purpose, a foreseeable and reproducible outcome, or neither? (Intended / Foreseeable / No / Unsure) | Art 5(1a)(a)(i)/(ii) |
| H | Provider only: safeguards that reliably prevent it and correct misuse? (Y/N/Unsure) | Art 5(1a)(a)(ii); recital 12 |
| I | Deployer only: used for the purpose of generating such material? (Y/N/Unsure) | Art 5(1a)(b); recital 12 |
| J | Explicit consent of the depicted person for that generation? (Y/N/NA) | Art 5(1)(ba) |
| K | Still offered or in use on or after 2 Dec 2026? (Y/N) | Drives STOP vs IN SCOPE |
| L | Art 5(1)(ba) 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: CAN IT PRODUCE REALISTIC INTIMATE MATERIAL OF AN IDENTIFIABLE PERSON? (ART 5(1)(ba); ART 5(1b))",IF(E2<>"Y","EXCEPTION OR OUT OF SCOPE: NO REALISTIC INTIMATE MATERIAL OF IDENTIFIABLE PERSONS ON THESE ANSWERS (ART 5(1)(ba); ART 5(1b))",IF(F2="Provider",IF(OR(G2="Unsure",G2=""),"CHECK BY HAND: INTENDED OR FORESEEABLE OUTCOME UNCLEAR (ART 5(1a)(a))",IF(G2="Intended",IF(AND(K2="Y",Settings!$B$1>=DATE(2026,12,2)),"STOP: IN SCOPE OF ART 5(1)(ba) BAN — SWITCH OFF OR FIX (APPLIES SINCE 2 DEC 2026)","IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026"),IF(G2="Foreseeable",IF(OR(H2="Unsure",H2=""),"CHECK BY HAND: SAFEGUARDS UNCLEAR (ART 5(1a)(a)(ii))",IF(H2="Y","EXCEPTION OR OUT OF SCOPE: SAFEGUARDS CLAIMED (ART 5(1a)(a)(ii)) — KEEP EVIDENCE THEY RELIABLY PREVENT AND CORRECT MISUSE",IF(AND(K2="Y",Settings!$B$1>=DATE(2026,12,2)),"STOP: IN SCOPE OF ART 5(1)(ba) BAN — SWITCH OFF OR FIX (APPLIES SINCE 2 DEC 2026)","IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026"))),"EXCEPTION OR OUT OF SCOPE: NOT INTENDED OR FORESEEABLE ON THESE ANSWERS (ART 5(1a)(a))"))),IF(F2="Deployer",IF(OR(I2="Unsure",I2=""),"CHECK BY HAND: PURPOSE OF USE UNCLEAR (ART 5(1a)(b))",IF(I2<>"Y","EXCEPTION OR OUT OF SCOPE: NOT USED FOR THAT PURPOSE ON THESE ANSWERS (ART 5(1a)(b))",IF(J2="Y","CHECK BY HAND: EXPLICIT CONSENT CLAIMED (ART 5(1)(ba)) — KEEP PROOF FOR EACH PERSON AND EACH GENERATION",IF(AND(K2="Y",Settings!$B$1>=DATE(2026,12,2)),"STOP: IN SCOPE OF ART 5(1)(ba) BAN — SWITCH OFF OR FIX (APPLIES SINCE 2 DEC 2026)","IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026")))),"CHECK BY HAND: ROLE UNCLEAR — PROVIDER OR DEPLOYER? (ART 3(3)/(4))"))))))) | In scope / out of scope / check / STOP |
| M | Status vs today: =IF(A2="","",IF(LEFT(L2,9)="EXCEPTION","NO ART 5(1)(ba) ACTION ON THESE ANSWERS",IF(LEFT(L2,5)="CHECK","CHECK BY HAND",IF(LEFT(L2,4)="STOP","OVERDUE: PROHIBITED SINCE 2 DEC 2026 ("&(Settings!$B$1-DATE(2026,12,2))&" DAYS) — STOP NOW",IF(Settings!$B$1<DATE(2026,12,2),IF(K2="Y","FIX OR SWITCH OFF BEFORE 2 DEC 2026: "&(DATE(2026,12,2)-Settings!$B$1)&" DAYS LEFT","DO NOT LAUNCH UNLESS FIXED: BAN FROM 2 DEC 2026 ("&(DATE(2026,12,2)-Settings!$B$1)&" DAYS LEFT)"),"DO NOT SWITCH ON: BANNED SINCE 2 DEC 2026 (ART 113(a) AS AMENDED)"))))) | Vs Settings!B1 and 2 Dec 2026 |
| N | To-do list: =IF(A2="","",TEXTJOIN("; ",TRUE,IF(C2<>"Y","not an AI system on these answers (Art 3(1)); keep the note",""),IF(AND(C2="Y",D2<>"Y"),"re-check whether it is placed on the market, put into service or used in the EU, or its output is used in the EU (Art 2)",""),IF(AND(C2="Y",D2="Y",OR(E2="Unsure",E2="")),"test whether the tool can produce realistic images, video or audio of an identifiable person's intimate parts or sexually explicit activity; edits that do not increase exposure are not manipulation (Art 5(1b))",""),IF(AND(LEFT(L2,5)="CHECK",F2<>"Provider",F2<>"Deployer",E2="Y"),"decide whether the client is the provider (develops or places the tool on the market under its own name, Art 3(3)) or a deployer (uses it, Art 3(4))",""),IF(AND(F2="Provider",E2="Y",OR(G2="Foreseeable",G2="Unsure",G2="")),"list the safeguards in place (e.g. refusal training, prompt guardrails, output filters, usage restrictions, abuse detection, notice-and-action; recital 12 of Reg (EU) 2026/1744) and the test that shows they reliably prevent the outcome and correct reported misuse",""),IF(AND(F2="Deployer",E2="Y",I2="Y",J2="Y"),"keep proof of freely-given, specific, informed, unambiguous and explicit consent of each depicted person for that generation (Art 5(1)(ba)); GDPR still applies",""),IF(AND(F2="Deployer",E2="Y",I2="N"),"lawful use for other purposes is not banned for deployers, and accidental output is not covered (recital 12); keep a written use policy that forbids this use",""),IF(OR(LEFT(L2,4)="STOP",LEFT(L2,8)="IN SCOPE"),"deployer: stop this use; provider: remove the function or add safeguards that reliably prevent it, before 2 Dec 2026 (Art 113(a) as amended by Reg (EU) 2026/1744); Art 5 breaches can attract fines up to EUR 35 000 000 or 7% of worldwide turnover (Art 99(3); for SMEs the lower cap, Art 99(6))",""),"record who decided, why and the date")) | Cites each article |
Step 2: what the test rows returned
Tested in LibreOffice Calc on 9 Nov 2026 with Settings!B1 fixed at 9 Nov 2026, on twelve made-up rows (no real business) plus an empty row. All twelve gave the intended result; the empty row stayed blank; no formula errors.
| Example | Art 5(1)(ba) result | Status vs today |
|---|---|---|
| 1: 'undress' photo app sold to EU users | IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026 | FIX OR SWITCH OFF BEFORE 2 DEC 2026: 23 DAYS LEFT |
| 2: paper photo retouch service, no AI | EXCEPTION OR OUT OF SCOPE: NOT AN AI SYSTEM ON THESE ANSWERS | NO ART 5(1)(ba) ACTION ON THESE ANSWERS |
| 3: general image generator with no nudity filter | IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026 | FIX OR SWITCH OFF BEFORE 2 DEC 2026: 23 DAYS LEFT |
| 4: general image generator with tested refusal + output filters + abuse reporting | EXCEPTION OR OUT OF SCOPE: SAFEGUARDS CLAIMED (ART 5(1a)(a)(ii)) — KEEP EVIDENCE THEY RELIABLY PREVENT AND CORRECT MISUSE | NO ART 5(1)(ba) ACTION ON THESE ANSWERS |
| 5: US-only app, no EU users or output | EXCEPTION OR OUT OF SCOPE: NOT IN EU SCOPE ON THESE ANSWERS (ART 2) | NO ART 5(1)(ba) ACTION ON THESE ANSWERS |
| 6: agency uses image AI for product shots only | EXCEPTION OR OUT OF SCOPE: NOT USED FOR THAT PURPOSE ON THESE ANSWERS (ART 5(1a)(b)) | NO ART 5(1)(ba) ACTION ON THESE ANSWERS |
| 7: staff member asked to 'nudify' a colleague photo with a chatbot | IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026 | FIX OR SWITCH OFF BEFORE 2 DEC 2026: 23 DAYS LEFT |
| 8: boudoir studio claims written consent for AI edits | CHECK BY HAND: EXPLICIT CONSENT CLAIMED (ART 5(1)(ba)) — KEEP PROOF FOR EACH PERSON AND EACH GENERATION | CHECK BY HAND |
| 9: background/colour edit that does not increase exposure | EXCEPTION OR OUT OF SCOPE: NO REALISTIC INTIMATE MATERIAL OF IDENTIFIABLE PERSONS ON THESE ANSWERS (ART 5(1)(ba); ART 5(1b)) | NO ART 5(1)(ba) ACTION ON THESE ANSWERS |
| 10: software house unsure if it is provider or deployer | CHECK BY HAND: ROLE UNCLEAR — PROVIDER OR DEPLOYER? (ART 3(3)/(4)) | CHECK BY HAND |
| 11: planned 'nudify' feature, not launched | IN SCOPE OF ART 5(1)(ba) BAN FROM 2 DEC 2026 | DO NOT LAUNCH UNLESS FIXED: BAN FROM 2 DEC 2026 (23 DAYS LEFT) |
| 12: avatar app, nobody tested intimate outputs | CHECK BY HAND: CAN IT PRODUCE REALISTIC INTIMATE MATERIAL OF AN IDENTIFIABLE PERSON? (ART 5(1)(ba); ART 5(1b)) | CHECK BY HAND |
Step 3: checklist
- Is an AI system involved (Article 3(1)), and is it placed, put into service or used in the EU, or its output used there (Article 2)?
- Can it produce realistic images, video, audio or similar material of an identifiable person's intimate parts or sexually explicit activity? Recital 12 says "realistic" means a credible real-life depiction of face, voice or body and excludes cartoonish or physically impossible depictions.
- Is the client the provider or a deployer of the tool (role check sheet)?
- Provider: is that output the intended purpose, or a foreseeable and reproducible outcome? If foreseeable, which safeguards prevent it, how were they tested, and how is reported misuse corrected (Article 5(1a)(a))?
- Deployer: is there a written rule that forbids using any tool for this purpose (staff AI use policy template)?
- Where consent is relied on, is there proof that each depicted person gave freely-given, specific, informed, unambiguous and explicit consent for that generation (Article 5(1)(ba)), handled under the GDPR?
- Is the fix done before 2 December 2026, and who signed it off and when?
Step 4: what to tell the client (template)
Hello [name], from 2 December 2026 the EU AI Act bans AI systems that generate or manipulate realistic intimate images, video or audio of an identifiable person without that person's explicit consent (Article 5(1)(ba), added by Regulation (EU) 2026/1744). For your [tool], our one-row check returned [result]. Before 2 December 2026 please [remove the function / document and test the safeguards / adopt the attached use rule], and send us who approved it and the date. This is a regulatory checklist, not legal advice; for any borderline row please take a written legal opinion.
Sources read
Regulation (EU) 2026/1744 (Digital Omnibus on AI), CELEX 32026R1744: recitals 10, 11, 12 and 14; Article 1 point amending Article 5 (new Article 5(1)(ba), (bb), 5(1a), 5(1b)); the amendment to Article 113(a). Regulation (EU) 2024/1689 (AI Act), CELEX 32024R1689: Articles 2, 3, 99 and 113. EUR-Lex refused automated fetches today (HTTP 202/403), so we used our saved copies of the official EUR-Lex text and checked the quoted wording against them on 9 Nov 2026. Not read for this guide: Commission guidelines on prohibited practices as they may be updated for points (ba)/(bb); national criminal law on intimate images; Directive 2011/93/EU in full; any case law. Our reading, flagged: the column grades, treating a consent claim and unclear role as CHECK BY HAND, and treating documented safeguards as out of scope only with kept evidence.
Frequently asked questions
What does Article 5(1)(ba) of the AI Act ban?
It bans placing on the market, putting into service or using an AI system that generates or manipulates realistic images, videos, audio or similar material of an identifiable natural person's intimate parts, or of that person engaged in sexually explicit activities, without that person's freely-given, specific, informed, unambiguous and explicit consent for that generation or manipulation. The point was inserted into Article 5(1) of Regulation (EU) 2024/1689 by the Digital Omnibus, Regulation (EU) 2026/1744.
When does the intimate deepfake ban apply?
From 2 December 2026. Article 113(a), as amended by Regulation (EU) 2026/1744, keeps 2 February 2025 for Chapters I and II but makes Article 5(1) points (ba) and (bb) and Article 5(1a) and (1b) apply from 2 December 2026.
Is a general image generator banned just because misuse is possible?
No, not on that ground alone. For providers the ban covers systems whose intended purpose is that output, or where it is a reasonably foreseeable and reproducible outcome without significant technical modification and the system lacks reasonable and adequate safeguards to reliably prevent it and to correct observed or reported misuse (Article 5(1a)(a)). Recital 12 lists examples of safeguards such as refusal training, runtime prompt guardrails, output filters, usage restrictions, abuse detection and notice-and-action mechanisms.
Is a business that only uses an image tool caught?
A deployer is caught only where it uses the system for the purpose of generating or manipulating such material (Article 5(1a)(b)). Recital 12 says lawful use for other purposes and accidental generation are not covered.
What are the fines?
Breaches of Article 5 can attract fines of up to EUR 35 000 000 or 7% of worldwide annual turnover, whichever is higher; for SMEs the lower of the two applies (Article 99(3) and 99(6)).
Related: Article 5 prohibited practices list · AI Act dates after the Digital Omnibus · Article 99 fines and the SME lower-amount rule.
Keep one register of every client's AI tools and their Article 5 answers: the EURO AI Register has a free edition.
Unadhi Guides are free, independent explainers. We correct errors as soon as we find them: write to mininglife7@gmail.com.