01 / Field guide
The useful definition is about consent
Deepfake porn is sexually explicit or intimate-looking media in which a recognisable person’s likeness has been digitally created, replaced or manipulated. The depicted act did not happen merely because a convincing file exists. Yet the person can still experience harassment, workplace consequences, family distress, stalking or extortion.
A deepfake porn video may involve face replacement, synthetic motion, altered audio or a fully fabricated scene. A still image and a video can require different matching and reporting techniques, but the central questions remain the same: is the person identifiable, did they consent to the intimate depiction, where is it hosted and how is it being distributed?
The query deepfake porn ai often mixes tools, hosting sites, news reports and removal resources into one search result. Store availability, a “for entertainment” label or an 18+ notice does not establish lawful consent. This guide separates provider claims from platform policies, regulator findings and independent measurements.
- False does not mean harmless. The subject’s face, name and social relationships are real even when the depicted body or event is fabricated.
- Removal has layers. A file can be removed from a host, a post can be removed from a platform, and a result can be removed from a search engine. These are separate actions.
- Evidence should be proportionate. Preserve URLs, context, threats and report numbers. Avoid making new copies of the intimate material merely to prove that it exists.
SEARCH LANGUAGE COVERED — used for risk identification, not promotion
deepfake porndeepfake porn videodeepfake porn ai02 / Original micro-study
Four public reporting systems, six practical questions
We reviewed the public help and policy pages for Google Search, X, Reddit and StopNCII on 3 August 2026. Each system received one point when we could locate a clear answer to six questions a targeted person is likely to ask.
21 of 24 disclosures were located. All four systems explicitly covered synthetic or faked intimate material. Three described representative reporting, three described duplicate or repeat protection, and three showed a report outcome or case-status route.
Scoring method. The six binary checks were: synthetic imagery explicitly covered; a direct report route described; an authorised representative or person reporting for someone else recognised; duplicate/reappearance handling documented; status or outcome information documented; and the system’s scope limitation plainly stated. “Not found” is not proof that a feature does not exist internally.
Important limit. This is a documentation audit, not a stopwatch test. We submitted no intimate imagery, impersonated no victim and measured no response time. X’s outcome point uses its US Take It Down Act page, so the geographic limitation is carried into the table note rather than hidden.
| System | Synthetic named |
Direct route |
Report for another |
Duplicate protection |
Status / outcome |
Scope clear |
Score |
|---|---|---|---|---|---|---|---|
| Google Search | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | 6/6 |
| X | ✓ | ✓ | ✓ | — | ✓* | ✓ | 5/6 |
| ✓ | ✓ | ✓ | ✓ | — | ✓ | 5/6 | |
| StopNCII | ✓ | ✓ | — | ✓ | ✓ | ✓ | 5/6 |
03 / Reporting route review
Use the right system for the right layer
A high disclosure score does not make every system interchangeable. One removes discovery, one moderates posts, and one shares privacy-preserving fingerprints with participating platforms.
Google’s help page explicitly accepts requests about fabricated sexual imagery when the requester is identifiable, falsely depicted and the content was distributed without consent. The subject or a representative can report. Its newer Image Search flow can include future duplicates and related searches, with status visible through Results about you.
Limit: delisting changes Google results; it does not delete the original page or file from the host.
X’s global non-consensual nudity policy expressly includes face superimposition and digital manipulation. It documents both in-app and desktop reporting, permits up to five posts in one report and recognises an authorised representative such as a lawyer for cases requiring the depicted person’s context.
Limit: the reviewed global page does not describe proactive duplicate matching. Its separate 48-hour outcome process is specific to the US Take It Down Act.
Reddit Rule 3 covers media created or posted without permission, including synthetic “lookalike” and deepfake material. Its April 2026 help page instructs users to report the exact post or threat through the Non-Consensual Intimate Media category and states that Reddit uses StopNCII.
Limit: the public help pages we reviewed do not set out a comparable request-status dashboard or response-time promise.
StopNCII creates a hash on the adult subject’s device; the image itself is not sent to the service. Participating platforms look for matching files, and a case number plus PIN can be used to check or withdraw the case. Its FAQ explicitly includes synthetic or deepfake intimate images.
Limit: only the depicted adult can submit. It covers participating public-facing platforms, not the entire internet or encrypted messages. Under-18 cases are directed to Take It Down.
04 / Ecosystem case files
Three interventions, three different pressure points
The deepfake economy is not one website. It depends on infrastructure, payment, search, hosting, social distribution and product design. These cases show why a single takedown is valuable but rarely sufficient.
MrDeepFakes closed after losing a critical provider
NPR reported that MrDeepFakes announced permanent closure in May 2025 after a critical service provider ended service and data loss made continued operation impossible. A USENIX Security study had measured more than 43,000 sexual deepfake videos and 1.5 billion views by November 2023.
Lesson: infrastructure withdrawal can interrupt a large marketplace quickly, but copies and successor communities can persist elsewhere.
Undress.cc faced age-assurance enforcement
Ofcom’s confirmation decision identified Itai Tech Ltd as operator and found that highly effective age assurance was absent during the relevant 2025 period. It imposed £50,000 for the age-assurance breach and £5,000 for failure to answer an information request; the service had geo-blocked UK IP addresses.
Lesson: an 18+ sentence is not the same as a regulator-accepted age-assurance system.
Grok on X triggered two formal investigations
Ofcom opened a UK investigation on 12 January 2026 following reports of Grok being used to create and share sexualised images. The European Commission opened a DSA investigation on 26 January. Both proceedings examine platform risk assessment and mitigation; opening an investigation is not a final finding of breach.
Lesson: safety must be designed before a high-reach image feature launches, not only added after mass misuse becomes visible.
05 / Current law snapshot
UK creation offences and US removal duties now coexist
The rules below are a dated orientation, not advice for a specific case. Jurisdiction, age, consent, intent and the location of the platform or person can change the analysis.
England and Wales
Section 138 of the Data (Use and Access) Act 2025 introduced offences for intentionally creating, or requesting the creation of, a purported intimate image of an adult without consent or reasonable belief in consent. The provision came into force on 6 February 2026.
- The Crime and Policing Act 2026 section 99 targets making, adapting, supplying or offering to supply purported intimate-image generators, subject to statutory definitions and defences.
- Sharing or threatening to share intimate images without consent is separately addressed by the Sexual Offences Act framework.
- Images involving children engage separate, especially serious offences. Do not retain or transmit suspected illegal child sexual abuse material.
United States platform process
The FTC began enforcing the notice-and-removal part of the TAKE IT DOWN Act on 19 May 2026. Covered platforms must provide a clear process and remove qualifying non-consensual intimate images and known identical copies within 48 hours of a valid request.
- The FTC guidance explicitly includes “digital forgeries” created or altered by software, an app or AI.
- A reporter does not need to be an account holder, and platforms are encouraged to provide a trackable request number.
- If a covered platform has no working process or fails to act, the FTC now provides a separate compliance complaint route.
06 / Evidence-first response
Preserve enough, report precisely, escalate safely
A targeted person should not have to become a forensic investigator. The aim is a small, reliable evidence package that can travel through platform reports, police contact and qualified legal advice without amplifying the image.
Capture context
Record the exact page, post and media URLs; account handle; visible caption; date, time and time zone; threats; and payment identifiers. Keep the first screenshot unedited and make a separate redacted copy for ordinary correspondence.
Work layer by layer
Report the post or host first. Then report search results, mirrors and reposts separately. Ask for removal, duplicate suppression where available, preservation of relevant logs and a case or confirmation number.
Prioritise safety
Do not pay an extorter or send more images. In the UK use 101 for non-emergency police advice and 999 for immediate danger. A solicitor or specialist image-abuse service can coordinate difficult multi-platform cases.
- Exact page, post and direct media URL
- Account name, profile URL and platform
- Date, time and time zone observed
- Threats, demands and payment details
- Non-explicit screenshot of page context
- Report category and confirmation number
- Outcome email and follow-up deadline
- Chronology of every repost or mirror
07 / Practical safety
Seven choices that reduce confusion and delay
No privacy setting can guarantee prevention once a clear image is public. These habits are about limiting high-quality source material and making response faster.
- Reduce public originalsReview old albums, public cloud shares and tagged photographs. Where practical, avoid publishing unnecessarily high-resolution, front-facing images.
- Approve tagsRequire review before tagged posts appear on your profile, and ask close contacts not to publish identifying photographs without checking.
- Secure recoveryUse unique passwords, a password manager, multi-factor authentication and offline recovery codes for email, social and cloud-photo accounts.
- Separate reporting emailUse a safe contact address that is not the login for your primary accounts when a platform asks for ongoing report correspondence.
- Keep a case ledgerOne spreadsheet or note with URLs, dates, report IDs and outcomes prevents repeated work and gives advisers a reliable chronology.
- Do not test generatorsUploading any real person’s image to “see if it works” creates privacy and consent risks. This page never requires a test upload.
- Protect wellbeingAsk a trusted person to monitor reposts or handle routine forms. Repeatedly viewing abusive imagery can compound harm without improving the evidence.
Device and browser accessibility
The page uses a system sans-serif at 17 pixels with a 1.7 line height and a reading measure of approximately 66 characters. Text remains left-aligned and no content block has a fixed height. The layout tolerates increased letter, word and paragraph spacing without clipping or text-on-text overlap.
The core guide is semantic HTML and remains readable when JavaScript is unavailable; only the situation switcher, focus mode, print shortcut and copy button are enhanced. Current Chrome, Edge, Firefox and Safari should receive the complete layout across Windows, macOS, Android and iOS. On narrow screens, the audit table scrolls horizontally rather than compressing its labels.
08 / Direct evidence
Sources readers can inspect
Provider statements are labelled as provider statements; investigations are not described as final judgments. No source link below points to an explicit generator. Every outbound link carries nofollow and sponsored attributes at the publisher’s request.
- USENIX Security 2025 — Characterizing the MrDeepFakes Sexual Deepfake MarketplacePeer-reviewed measurement of videos, views, marketplace participants and creation-community behaviour.
- Google Search Help — Remove personal sexual contentEligibility, representative reporting, duplicate handling and search-removal scope.
- X Help — Non-consensual nudity policyManipulated imagery examples, reporting steps, representative reporting and enforcement language.
- Reddit Help — Rule 3 and synthetic intimate mediaPublic rule covering faked, lookalike and AI-generated sexual media.
- StopNCII — How it worksOn-device hashing, participating-platform matching and scope limitations.
- Ofcom — Investigation into X over Grok sexualised imageryOpened 12 January 2026 and recorded as ongoing in the reviewed update.
- US FTC — Complying with the TAKE IT DOWN ActEffective 19 May 2026 platform-process guidance, identical-copy duties and enforcement context.
- UK legislation — Section 138 commencement instrumentOfficial instrument bringing the creation and request offences into force on 6 February 2026.
- Tech Transparency Project — Nudify apps in mobile storesJanuary 2026 AppMagic-based count used for the 55 Google Play and 47 Apple App Store indicator.
- Elliptic — Analysis of deepfake undresser services2025 industry analysis covering more than 120 services and the payment infrastructure around them.
- NPR — MrDeepFakes announces permanent closureContemporaneous reporting on the loss of a critical service provider and the site’s shutdown statement.
- Ofcom — Itai Tech Ltd confirmation decisionPrimary enforcement decision identifying the operator, relevant dates, geo-block and financial penalties.
- European Commission — Grok and X DSA investigationOfficial notice opening the investigation on 26 January 2026; it is not a final infringement decision.
- X Help — US Take It Down Act processUS-specific request route, 48-hour language and outcome notification used in the audit note.
- Reddit Help — Reporting intimate-image abuseApril 2026 platform instructions and Reddit’s documented use of StopNCII matching.
- StopNCII — Frequently asked questionsEligibility for synthetic imagery, self-submission, case status and participating-platform limits.
- Crime and Policing Act 2026 — explanatory notesOfficial explanation of section 99 and the offence concerning purported intimate-image generators.
09 / Publisher transparency
Identity is verified; expertise is evidenced separately
The corporate record identifies the publisher and aligns in part with video and consultancy work. It does not turn a company filing into legal, clinical or forensic accreditation.
MW INTELLIGENCE LIMITED
- Company no.
- 17149132
- Status
- Active
- Type
- Private limited company
- Incorporated
- 10 April 2026
- Office
- 1 Parkshot, Richmond, Surrey, England, TW9 2RD
- SIC 59112
- Video production activities
- SIC 70229
- Management consultancy activities other than financial management
- Registry
- Companies House overview
Editorial accountability
Deepfake XXX: Evidence & Removal Desk uses a dated methodology, direct policy pages, primary legislation, peer-reviewed research and explicit limits. Provider statements and regulator findings are not merged. The 88% figure can be reproduced from the six published criteria and four named systems.
Companies House itself warns that it does not check the accuracy of filed information. The register is therefore used only to verify the public corporate record shown here. Readers should obtain qualified legal advice for individual cases and professional mental-health support when needed.
10 / FAQ
Five concise answers
Is deepfake-xxx.net a deepfake porn generator?
No. It is an independent evidence and removal guide. It accepts no images, creates no media and provides no links to generators.
Does removing a result from Google delete the original file?
No. Search removal limits discovery in Google Search. The host or platform must be contacted separately to remove the source file, and reposts may require separate reports.
Can StopNCII be used for a synthetic image?
StopNCII says an adult depicted in a nude or semi-nude synthetic image can hash it if they have access to the file. The file remains on the device, but matching works only on participating public-facing platforms.
Is creating non-consensual synthetic intimate imagery illegal in England and Wales?
Section 138 of the Data (Use and Access) Act 2025 created offences covering intentional creation or requesting creation of a purported intimate image of an adult without consent or reasonable belief in consent. It came into force on 6 February 2026. Specific cases require qualified advice.
What does the 88% reporting audit measure?
It records 21 of 24 selected public-process disclosures located across Google Search, X, Reddit and StopNCII on 3 August 2026. It is a documentation audit, not a measure of response speed, moderation quality or legal compliance.