The Explosion of Deepfake Porn: AI Tools to Detect and Deter Manipulation

1. Deepfake porn generated by AI is exploding, with millions of victims ranging from everyday people to children and celebrities.
2. Tools like digital watermarks and “poison pills” aim to detect and protect against damaging deepfakes.
3. Regulation is trying to catch up, but faces free speech issues around personal creation and viewing of deepfakes.
4. Pressuring tech companies to introduce friction and obstacles can help combat spread, but completely stopping deepfakes may not be possible.
5. The trauma experienced by victims of nonconsensual deepfakes is very real, even if the content itself is fake.
6. Criminalizing creation and distribution is an important deterrent, even if difficult to fully enforce.

The risks of artificial intelligence can seem overwhelming. For every benefit AI provides, there is an adverse use. One major problem is deepfakes: videos, images, or audio generated by AI that mimic a victim saying or doing something that never happened.

Deepfakes can be designed to superimpose a likeness onto real footage, while others are completely computer-generated. A 2019 study by Deep Trace found 96% of deepfake videos were pornographic. Researcher Henry Ajder says while the figure may have changed, the volume of pornographic deepfakes has exploded into the millions.

Most victims are everyday people, but children and celebrities are also targeted. While the content is fake, the trauma for victims is real. In 2021, a British teenager killed herself after deepfake pornographic images of her were shared in a Snapchat group.

Fighting Back Against Deepfake Porn: Detection, Protection, and Regulation

As AI tools like Dall-E and Stable Diffusion become more accessible, it’s easier for people with little technical skill to create deepfakes. Last month, deepfake porn of Taylor Swift circulated online. Elon Musk banned searches of her on X to combat this.

There are tools and methods that can help protect against AI manipulation:

Deepfake Detection
Digital watermarks clearly label AI-generated content to raise awareness and help platforms remove damaging fakes. Google, Meta, and OpenAI plan to add visual watermarks and metadata revealing a photo’s history.

Platforms like Sensity alert users via email when media has telltale AI fingerprints. But even obvious fakes may still victimize subjects psychologically.

‘Poison Pills’
Defensive tools add imperceptible signals to images that corrupt them when fed to AI systems. For example, Nightshade adds pixels that confuse AI models but leave images intact for humans. This protects artists’ IP and personal photos.

Regulation
Over 10 states have legal protections for deepfake victims. Recent high-profile cases have increased federal pressure. The FCC banned AI-generated robocalls after hoax calls of Joe Biden. A new bill would let victims sue deepfake creators.

But legislation faces free speech issues. Some see private deepfake creation as akin to a personal fantasy. If it’s not shared, has harm occurred? This has affected UK law, which bans distribution but not creation of deepfakes. Criminalization is still important to deter curious creators, argues Ajder.

Governments can also pressure search engines, AI developers, and social media platforms to introduce friction against deepfakes. After a celebrity scandal, India fast-tracked laws and pressed tech giants to prevent spread. Total removal may be impossible, but added obstacles help, says Ajder.

From the media observer’s perspective, the deepfake phenomenon represents a critical inflection point in the evolution of personal privacy and consent. As AI generation becomes more accessible, deepfakes foreshadow an impending reality where any individual’s identity can be seamlessly co-opted to create nonconsensual media.

The resulting psychological and emotional violation for victims is immense and threatens the very sanctity of self-ownership. Yet legislation wrestles with thorny questions around free speech and fantasy. This tension embodies society’s struggle to balance individual rights with preventing harm in a rapidly changing technological landscape.

For the media industry, the stakes are also high. Deepfakes directly attack public trust in information integrity. If compelling forgeries proliferate unchecked, how can citizens discern truth? Platform accountability will be scrutinized. And news outlets must grapple with inadvertently spreading manipulated content that may not be obviously fake.

Ultimately, the observer recognizes that while detection tools and regulations can combat deepfakes, they cannot eliminate the Pandora’s box of issues opened by AI generation technology. The challenge ahead is monumental. But solutions must be pursued to uphold personal consent and truth — the bedrocks of civil society — in the face of threats old laws never envisioned. The future remains unwritten, but the need for wisdom and nuance is more vital than ever.