Pornhub says it's banning AI-generated fake porn videos because they're “nonconsensual”, yet such “deepfake” videos are still available via site search results
In its search of a platform that will host fake porn videos of celebrities created with a machine learning algorithm …
Context & Ripple Effects
Pornhub's ban lands two weeks after Motherboard documented how FakeApp-style face-swap tools put convincing celebrity fake porn within reach of anyone with a laptop — so the policy is a response to a supply problem that had already gone mainstream. The site is also the same company that months earlier announced an AI facial-recognition system to identify and tag performers, meaning it owns detection technology purpose-built for exactly this content.
The gap between announcement and enforcement matters because the corpus shows this pattern repeating: years later, non-consensual material was still surfacing in machine learning training datasets (VICE's reporting on dataset persistence) and model-sharing sites like CivitAI were hosting generators trained on unconsented material. A stated ban that search results contradict is the earliest data point in that arc.
First-order effects
- Users searching Pornhub right now can still surface the very deepfake videos the company has declared nonconsensual, making the ban a policy statement rather than an enforced rule.
- Pornhub's own moderation credibility takes the hit: a platform that announced facial-recognition tagging in 2017 cannot claim technical inability as the reason fakes remain searchable.
Second-order effects
- Competing tube sites face pressure to match the ban publicly while facing the same detection burden privately — the announcement sets a norm that costs nothing to state and real money to enforce.
- Advertisers and payment processors, who judge platforms on moderation follow-through rather than press releases, get fresh evidence that stated policy diverges from search-results reality.
Third-order effects
- If the pattern holds — announce, under-enforce, get caught by journalists — synthetic-media moderation shifts from voluntary platform policy toward external verification, whether via regulators, payment rails, or third-party detection audits.
- The deeper structural issue the corpus keeps returning to is likeness governance: once non-consensual imagery enters datasets and models, removal at any single host is insufficient, pushing responsibility upstream to the tools and training data.
The trend: Platform policies on synthetic sexual media are converging on a cycle where announcements outpace enforcement, and the gap between the two becomes the story regulators and journalists chase.