In late 2017, the internet witnessed a dark milestone in the evolution of artificial intelligence: the emergence of pornographic deepfakes featuring celebrities like Scarlett Johansson, Gal Gadot, and Taylor Swift. What began as a niche experiment on obscure online forums exploded into a viral phenomenon, thrusting deepfake technology into the spotlight for all the wrong reasons. These AI-generated videos—where the faces of famous women were seamlessly superimposed onto adult film actors—weren’t just a technical marvel; they were a chilling breach of privacy, consent, and ethics. Dubbed one of the earliest and most infamous deepfake scandals, this “krass” case sparked outrage, legal debates, and a reckoning over the unchecked power of AI. Let’s dive into how it happened, its fallout, and what it revealed about the double-edged sword of synthetic media.

The Setup: A Sinister Experiment Goes Viral

It started quietly in the shadowy corners of Reddit, where a user known as “deepfakes” (later inspiring the term itself) unveiled a disturbing creation. Using open-source AI tools, this individual had taken publicly available footage of celebrities—movie clips, interviews, red-carpet appearances—and melded it with adult film content. The result? Videos showing Scarlett Johansson, Gal Gadot, Natalie Portman, and others appearing to perform in explicit scenes they never consented to. The technology wasn’t flawless—there were glitches in lighting and occasional uncanny distortions—but it was convincing enough to fool casual viewers and ignite a firestorm.

The process was disturbingly simple. The creator fed hours of celebrity footage into a Generative Adversarial Network (GAN), a type of AI where two neural networks compete: one generates fake content, the other critiques it until the output improves. Paired with adult video frames, the AI mapped the stars’ facial features—eyes, lips, jawlines—onto the bodies of pornographic actors. Lip-syncing wasn’t even necessary; the focus was on visual deception. Within weeks, these videos spread beyond Reddit to platforms like Twitter and Pornhub, racking up millions of views. What began as a tech geek’s proof-of-concept became a global scandal, fueled by voyeurism and the internet’s insatiable appetite for controversy.

How Did They Pull It Off?

The 2017 deepfake scandal didn’t require a supercomputer or a Hollywood budget—just a decent laptop and free software. Tools like FakeApp, built on TensorFlow and other open-source frameworks, had democratized deepfake creation by late 2017. Anyone with basic coding skills could download it, gather training data (e.g., celebrity photos or videos), and let the AI do the heavy lifting. For high-profile targets like Johansson or Gadot, the data was abundant—think movie trailers, YouTube interviews, or paparazzi shots. The more footage, the better the AI could mimic subtle expressions and head movements.

The process took hours or days, depending on the hardware, but the barriers were low. A typical workflow involved extracting frames from a source video (the celebrity), training the GAN on those frames, and then overlaying the generated face onto a target video (the adult content). The results weren’t perfect—edges sometimes blurred, and lighting mismatches betrayed the fakery—but they were good enough to deceive untrained eyes. The creator bragged about the ease of it all, sharing tutorials that spawned a wave of copycats. By early 2018, the trend had metastasized, with DIY deepfakes popping up across the web, targeting not just celebrities but private individuals too.

The Aftermath: Outrage and Reckoning

The backlash was swift and fierce. Celebrities like Scarlett Johansson spoke out, calling the videos “demeaning” and “a perversion of technology.” Johansson, whose likeness was among the most exploited, later reflected on the futility of fighting back: “The internet is a vast wormhole of darkness that eats itself.” Social media platforms scrambled to act—Reddit banned the original deepfakes subreddit in February 2018, citing non-consensual content violations, while Pornhub and Twitter purged thousands of uploads. But the genie was out of the bottle; mirrors and archives kept the videos alive on less regulated corners of the web.

Public reaction oscillated between horror and fascination. News outlets ran exposés, decrying the ethical lapse, while some tech enthusiasts marveled at the AI’s capabilities. Victims faced a double blow: not only had their images been hijacked, but the viral spread amplified their humiliation. Johansson, already a frequent target of privacy invasions (e.g., hacked nude photos in 2011), became a symbol of the scandal’s human cost. Legal recourse proved elusive—U.S. laws in 2017 lacked specific provisions for deepfakes, and tracking anonymous creators across jurisdictions was a nightmare.

The scandal also hit the adult industry. Performers whose bodies were used as “templates” without consent voiced outrage, highlighting a layered violation: celebrities weren’t the only victims. Meanwhile, the tech community faced soul-searching. AI researchers condemned the misuse, but the tools’ open-source nature made control impossible. The episode turned “deepfake” into a household term, synonymous with danger rather than innovation.

Why This Case Stands Out

This 2017 scandal is “krass” for its pioneering role and raw audacity. It wasn’t the first synthetic media—Photoshop had long faked images—but it was the first to weaponize AI video at scale against unwilling targets. Unlike later financial scams (e.g., the $25.6 million Hong Kong heist), this wasn’t about money; it was personal, visceral, and deliberately provocative. It exposed a new frontier of harm: non-consensual digital exploitation that anyone could replicate.

The celebrity focus amplified its impact. These weren’t obscure figures but A-listers with global followings, making the violation feel both intimate and universal. The ease of creation was equally shocking—unlike traditional film trickery, this required no studio, just a lone coder and a laptop. And while the tech was rudimentary compared to 2025 standards, it was a proof-of-concept that foreshadowed worse to come: from political deepfakes to targeted revenge porn against everyday people.

The ethical breach was glaring. Consent, already a flashpoint in the digital age, was obliterated. The scandal forced society to confront a grim reality: if celebrities with legal teams and public platforms couldn’t stop this, what chance did regular individuals have?

The Bigger Picture: A Pandora’s Box Unleashed

The 2017 celebrity deepfakes opened a floodgate. By 2018, the tech spread to political disinformation (e.g., the Zelenskyy fake) and financial fraud, but its roots in pornographic misuse shaped its early infamy. It revealed the dark side of AI’s democratization—tools meant for creativity became instruments of harm. The scandal also highlighted a legal lag: most countries had no framework to address synthetic media crimes, leaving victims in limbo.

Tech platforms faced a reckoning too. Their reactive bans couldn’t keep pace with the deluge, and algorithms struggled to detect early deepfakes. The incident spurred research into countermeasures—watermarking, detection AI—but also a race among bad actors to refine their craft. By 2025, deepfakes have grown more sophisticated, with real-time capabilities and near-perfect fidelity, but 2017 was the tipping point that showed what was possible.

Lessons Learned and What’s Next

The scandal taught the world that deepfakes weren’t a sci-fi fantasy—they were here, and they were dangerous. It pushed platforms to tighten policies and invest in detection, though enforcement remains spotty. For individuals, it underscored the risks of a digital footprint—every photo or video online is potential fodder for manipulation. Legal systems began to adapt, with some U.S. states passing anti-deepfake laws by 2019, but global consensus lags.

The future is daunting. Today’s deepfakes can impersonate anyone in real-time, from video calls to live streams, amplifying the threat beyond static porn clips. The 2017 case was a wake-up call, but it also normalized the tech’s misuse, inspiring a wave of revenge porn targeting non-celebrities. Defenses—better AI detectors, public awareness—evolve, but so do the tools of deception.

Conclusion: A Stain on AI’s Legacy

The 2017 celebrity porn deepfake scandal wasn’t the deadliest or costliest AI crime, but it was one of the most personal. It turned stars into unwilling avatars, exposed the fragility of privacy, and set a precedent for synthetic media’s dark potential. What started as a twisted experiment became a cultural flashpoint, forcing us to ask: if technology can rewrite reality this easily, what’s left of trust? Eight years later, we’re still grappling with the answer.

What’s your view? Can we tame deepfake misuse, or has the damage already been done? Share your thoughts below.