Protected identity for women in public life

You are being monetized.

Right now, your face and voice are generating revenue on platforms you have never visited. Drivon finds where it’s happening, documents it, and gets it taken down.

1.8M

Sexualized AI-generated images of women.
Created by Grok on X in nine days. New York Times analysis.1

Your likeness is being sold as a paid chatbot subscription. Your face is built into LoRA models that strangers download to generate images. Your voice is being cloned and posted to public libraries. Your name is being prompted into image generators — Grok alone produced 4.4 million images on X in nine days.1 People are making money from each of these things. The ones you know about are a small fraction of the ones happening.

Drivon finds them at the source — the chatbot configurations, the LoRA uploads, the voice clone libraries, the paid subscription accounts using your face — before they reach scale, and while there is still time to act.

Other services find finished deepfakes after they’ve spread. Drivon finds where you’re being monetized, replicated, and used in scams — before the spread happens.

Apply for protection →

02 — The state of the threat

What’s happening, right now

May 19, 2026

The TAKE IT DOWN Act is now law.Signed in May 2025, with full platform enforcement beginning May 19, 2026.2 Platforms have 48 hours to remove non-consensual intimate imagery — including AI-generated — after receiving notice from the depicted person. The FTC enforces, with civil penalties of up to $53,088 per violation.3

June 11, 2026

Federal seizures have begun.The DOJ and Department of Homeland Security seized CFAKE.com and SOCFAKE.com — the first enforcement action under the new law. The sites hosted hundreds of thousands of deepfake images of “politicians, first ladies of multiple countries, royalty, journalists, television presenters, athletes, entertainers, and others.” An operator was arrested in Nice, France.4

March 2026

Class-action lawsuits against xAI are active.Three girls filed a federal class action alleging Grok was used to generate child sexual abuse material from their photos. Tennessee teenagers filed a separate suit on March 16, 2026.5 Wallace Miller, Hagens Berman, and Ben Crump Law are handling synthetic-content cases requiring court-admissible evidence preservation.

Aug–Sept 2025

Chatbots impersonating real women without their consent operate on major platforms.In August 2025, Reuters investigators found that Meta hosted unauthorized AI chatbots impersonating multiple A-list female celebrities — Grammy-winning singers, Oscar-nominated actresses, and major media personalities among them — across Facebook, Instagram, and WhatsApp without the women’s consent. The bots claimed to be the actual celebrities, made sexual advances, generated photorealistic images of the women in lingerie, and invited users to in-person meet-ups. Meta removed about a dozen of them only after Reuters’ reporting.11 In September 2025, researchers separately documented an unauthorized Character.AI chatbot impersonating singer Chappell Roan engaging in inappropriate conversation with a test account registered as a teenager — one of 50 user-created bots logging 669 harmful interactions in 50 hours.12 On both platforms, the companies hosting the bots did not detect or remove them until reporters did.

35,000 models

Civitai hosts tens of thousands of LoRA models trained on real women.A peer-reviewed study identified nearly 35,000 publicly downloadable deepfake model variants, primarily on Civitai, downloaded almost 15 million times since November 2022. 96% target identifiable women.6 Photographer Jingna Zhang found a custom LoRA built from her work during a phone call with a colleague — she had no automated monitoring in place when it was uploaded.7

9.7M views

Platform takedown response is selective.A deepfake of the WNBA’s most popular player reached 9.7 million views on X. When a journalist reported it, X replied within 24 hours that the video would remain up.8 Documented refusals at this scale are the operating norm, not the exception.

03 — Where this comes from

The labs aren’t building this. Other companies are.

The first question clients ask is how this is possible. Aren’t AI companies supposed to refuse?

The major labs do refuse. Anthropic, OpenAI, and Google train their models to decline likeness generation and catch most attempts to work around it. None of the material Drivon finds is produced by Claude, GPT, or Gemini.

The harm comes from a smaller group of companies that made different choices.

01 — Open distribution

Companies that released their models with no restrictions on downstream use

Once a model is freely distributed, no enforcement is possible — there is no API to call, no policy to enforce, no company to notify. The major labs kept their models closed for this reason. Others did not.

02 — Permissive platforms

Platforms built specifically to host what the labs declined to serve

These platforms operate openly, monetize through tipping and subscriptions, and rely on the gap between what is technically legal to host and what is enforceable in practice.

03 — Under-moderated chat

Consumer chatbot platforms that allow user-created impersonating characters faster than they can be removed

A 2025 report from ParentsTogether and the Heat Initiative documented impersonating chatbots of named celebrities with hundreds of thousands of user interactions before any moderation action was taken.12

04 — Deepfake services

Subscription deepfake services operating from weak-enforcement jurisdictions

These are businesses with payment processors and marketing budgets, built specifically for non-consensual likeness work.

Drivon is downstream of all four. We find what they’ve done, document the monetization, and assemble the evidence needed to act on it.

04 — How it works

What Drivon does

All three layers run on Drivon OS, the platform that orchestrates detection, evidence capture, takedown coordination, and verification across every protected client.

01 — Detection

Detection at the source

Drivon scans the platforms where synthetic content is built, hosted, monetized, and shared — Civitai for LoRA models trained on your face, Character.AI and similar platforms for chatbots configured with your name, voice clone libraries where your voice has been cloned for resale, Telegram channels selling subscriptions to galleries built around your likeness, and platforms where your face has been attached to paid impersonation accounts collecting revenue. We find the artifacts and the revenue streams before they reach scale, not after the content has spread.

02 — Advocacy

Legal advocacy with investigative rigor

Every match is documented to evidentiary standard: archived to the Internet Archive, hashed for tamper-proof verification, screenshotted, and preserved with full chain of custody. We file under the TAKE IT DOWN Act, DMCA, right of publicity, and platform-specific impersonation policies. We escalate through FTC enforcement when platforms fail the 48-hour mandate. When civil action is warranted, we hand the evidence package to plaintiff-side counsel ready to file.

03 — Monitoring

Continuous monitoring under verified-channel access

Every protected client account is locked to verified social channels and a signed engagement agreement — not to a password that can be phished or guessed. We scan on schedule, report monthly, and escalate immediately when active threats emerge. The work doesn’t stop when a takedown succeeds. The same model gets re-uploaded; the same chatbot gets recreated. We monitor for that and respond.

05 — Live operations

What Drivon OS is finding

Drivon publishes anonymized summaries of what Drivon OS is catching in active operations. The format is intentionally narrative: enough specificity to convey what kind of activity Drivon is detecting and resolving, with identifying details removed to protect the people involved.

This week

Drivon OS enters production reporting in Q3 2026. Live summaries of detections, takedowns, and monetization channels disabled on behalf of protected clients will appear here on a weekly cadence as we go live.

Sample entries will read in the form of:

A popular American singer being replicated as multiple paid chatbots on a major AI platform — estimated thousands of conversations generated from a single bot before our notice was filed.
An Olympic gymnast’s likeness packaged into a downloadable model on a public image-generation platform, with over a thousand downloads, being used to generate content sold through a paid subscription channel.
A professional wrestler’s voice cloned and listed in a public voice library, documented being used in promotional content for a product line the wrestler has no involvement with.
A creator’s face attached to a fraudulent subscription account collecting recurring payments from misled followers, with documented redirect to a payment portal.

These are the kinds of finds Drivon’s detection layer surfaces. The actual entries will replace these examples when production reporting begins.

06 — Access control

Why we don’t accept self-service signups

Anyone can claim to be anyone online. A self-service signup with an email and password means a stalker can register an account in their victim’s name, see what Drivon has found, and use that information to escalate. That isn’t hypothetical — it’s how impersonation harassment works.

Drivon operates by application and verified-channel engagement only. Identity is confirmed through verified public channels — a post from the verified Instagram or X account, a confirmation from an agency contact at a verified domain, or a video call when neither is available. The engagement agreement is signed before any monitoring begins. The case file is locked to the verified identity, not to a password — managed inside Drivon OS, the platform that runs the detection, evidence capture, and takedown coordination for every protected client.

The application step also lets us decline cases we can’t serve well. We accept clients we can protect and decline ones we can’t, including cases requiring resources outside our scope. Where appropriate, we refer to partner organizations.

07 — About the founder

Drivon was built by an investigator

Drivon was founded by Jeremy Paige, a former licensed private investigator in Alberta, Canada (2023–2025). Working individual cases under NDA, he saw consistent patterns in how online harassment, impersonation, and synthetic-content-driven stalking actually operate — and a consistent gap between what platforms do when notified and what victims actually need.

Drivon is what he built to close that gap: investigative discipline applied with AI tooling, paired with active legal advocacy, operated under Vallon Studios Limited.

Every application, every engagement, every case is handled personally by the founder. Drivon has no junior staff, no sales team, and no offshore monitoring center. The person who reviews your application is the same person who oversees your case, signs your engagement agreement, and is accountable to you for the work.

Verifiable Credentials

Alberta private investigator license (2023–2025) is verifiable through Alberta Security Programs (Public Safety and Emergency Services), the body that licenses private investigators under the Security Services and Investigators Act. Direct verification: ssia.registrar@gov.ab.ca or 1-877-462-0791. Vallon Studios Limited is publicly verifiable through Alberta Corporate Registry.

Jeremy Paige on LinkedIn →

08 — Engagement

Apply for protection

Drivon accepts a limited number of Protected Clients per quarter. Founding cohort pricing is locked at $997 per month, annual contract, audit included — available for the first 10 clients, then standard rates apply.

The application takes about 10 minutes. Every submission is reviewed personally. We respond within 48 hours with either an acceptance, a decline with referral, or a request for additional information.

What we’ll ask for

Your name and the verified social channels you operate under
A brief description of what’s currently happening — chatbots, LoRAs, deepfakes, voice clones, ongoing harassment, or what you’re protecting against proactively
Your agent, manager, or legal representative’s verified contact, if applicable
Anything you’ve already documented — screenshots, URLs, news links — that helps us understand the scope

We don’t ask for original photos or biometric data at application. Verification happens through your verified public channels after acceptance, never before.

Apply for protection →

Other engagement types

Emergency Audit

$3,497

30-day intensive · no application gate

For clients in active crisis. When a deepfake is already at 9.7 million views and the platform won’t remove it,8 an Emergency Audit assigns a dedicated 30-day response — every available takedown filed, full evidence package built, civil case referral if viable.

Start Emergency Audit →

Standard Protected Client

$1,497 / mo

Annual contract · post founding cohort

Ongoing monitoring, takedown coordination, monthly reporting. A Protected Client retainer monitors continuously, so an upload like Jingna Zhang’s custom LoRA7 is found and challenged within hours of going public — not weeks or months later.

Apply for protection →

Legal Evidence Package

$3,997 · $7,997

Per case · standard or complex

For plaintiff-side firms handling synthetic-content litigation. With class actions against xAI now active and firms like Wallace Miller, Hagens Berman, and Ben Crump Law building case pipelines,5 a Legal Evidence Package delivers full chain of custody, hashed forensic captures, archived URLs, and expert-witness-grade documentation.

Contact for legal engagement →

Agency Channel

$1,497 / client / mo

$1,247 at three or more clients

For talent agencies and management companies covering multiple clients. Loti AI has signed partnerships with WME, CAA, Dolphin Entertainment, and Underscore Talent for consumer-facing protection.9 Drivon’s Agency Channel exists for agencies that want the same coverage at the boutique tier.

Contact for agency engagement →

Federation and Enterprise

From $35,000 / year

Custom quote · per organization

For sports federations, leagues, athletic departments, and rosters at scale. The WTA, ITF, USTA, and All England Lawn Tennis Club fund Signify Group’s Threat Matrix to monitor approximately 8,300 professional tennis players.10 The WNBA, LPGA, NWSL, and NCAA athletic departments — where AI-generated content of female athletes has been a documented problem — do not have equivalent coverage. A Federation engagement provides synthetic-content monitoring at the organization level, custom-scoped to roster size and threat surface.

Contact for enterprise engagement →

Citations