Characterizing the MrDeepFakes Sexual Deepfake Marketplace

Explore the rise and shutdown of MrDeepFakes, the largest sexual deepfake marketplace. Learn why tech alone can’t stop non-consensual AI porn and what policymakers, platforms, and AI engineers must do to prevent the next wave of abuse.
안수남's avatar
Dec 09, 2025
Characterizing the MrDeepFakes Sexual Deepfake Marketplace

Paper:

What Was MrDeepFakes?

MrDeepFakes was the world’s largest marketplace for non-consensual sexual deepfakes: pornographic videos made by digitally swapping faces into explicit content. It hosted over 43,000 videos, targeting more than 3,800 individuals, and drew 1.5 billion views before shutting down in 2025.

While it claimed to only allow celebrity content, research shows hundreds of videos featured non-famous, everyday people. A robust community of creators, buyers, and “deepfake artists” operated through a forum and video platform: commissioning, trading, and distributing sexual fakes at scale.

Why It’s More Than Just a Tech Problem

Sexual deepfakes are not harmless entertainment: they’re image-based sexual abuse. Victims often suffer from reputational damage, trauma, and loss of control over their likeness. With 95% of targets being women, the phenomenon reinforces existing gendered power dynamics.

Even worse, some creators justify their actions as “art” or “community contributions,” normalizing this abuse within online subcultures.

Rules That Didn’t Work

Although MrDeepFakes had formal rules, like banning underage targets, rape scenes, or non-celebrities they were poorly enforced:

  • 14% of targets were not public figures

  • Over 1,000 videos included rape or humiliation themes

  • Requests for underage deepfakes were not immediately removed

This shows that self-regulation is ineffective when profit and anonymity dominate.

Why Tech Alone Can’t Solve It

Community members skillfully bypassed technical barriers:

  • They used open-source tools like DeepFaceLab

  • Trained models on free cloud GPUs like Google Colab

  • Evaded detection using renamed files or alternate platforms

Tech companies banning deepfakes simply led users to find workarounds. Adversarial dynamics mean creators are always a step ahead.

MrDeepFakes is gone, but the tech and the harm remain. Stopping the next version will require collaboration across tech, policy, and society. The lesson? AI progress must be matched by ethics, safeguards, and accountability.

Share article

Best Online LaTeX Editor, Murfy