Forget, for the moment, your fears that AI is draining our dams, making us dumb, and plotting our planetary overthrow. Because right now it’s posing a more immediate albeit less existential threat: It’s legit undressing us.
Popular so-called “nudify” apps use AI to turn photos of clothed humans into images of those same humans in their birthday suits. According to CBS News, more than 700 million of these apps have been downloaded worldwide, and one in 10 teens say they know someone who’s been a target. Even Papua New Guinea is trying to legislate this, you guys. It’s everywhere.
To be fair, Photoshop made this pervy practice possible decades ago. But you needed a smidgeon of skill. AI makes it easy — obscenely easy, you might say — to produce not only raunchy images of real people, but (gulp) deepfake X-rated videos, too.
Are we surprised that one of humanity’s most powerful inventions is being aimed up skirts? Maybe not. As a species, we’re disappointingly predictable, and given the opportunity to virtually undress a favorite actor or singer … tech nerds gonna tech nerd.
But these disrobing apps are also being used against everyday folks — and maliciously. Men have been arrested for creating and distributing explicit images of un-consenting ex-wives and girlfriends as a form of breakup revenge. And cyberbully schoolkids are using the tools on classmates. Perhaps most disturbing, the American Bar Association says that AI-generated kiddie porn has “expanded with alarming speed.”
As usual, lawmakers are struggling to catch up with the technology. The Take It Down Act requires online platforms to remove any sexual images published without the subject’s consent within 48 hours of notification. And Rep. Alexandria Ocasio-Cortez (D, NY) — herself a victim of such deepfakes — is trying to pass a law that lets victims sue the creators of these deepfakes.
Part of the problem with regulating this mess may be that, frankly, we’re still wrestling with what exactly we’re looking at. Anyone with a moral compass knows it’s wrong. We even know why it’s wrong. But specifically, logically, and philosophically: How is it wrong? Is it theft? Is it libel? Is it an “invasion of privacy” if the image doesn’t reveal the subject’s real body?
An AI engineer I know has given this a lot of thought.
“Calling the image ‘fake’ explains how it was produced. It tells us nothing about the reality of the violation,” says Jeremiah Clark, chief technology officer of an AI image and video creation company. “Creating nonconsensual sexual or degrading synthetic media of a real person is image-based abuse.”
Our online lives can no longer be considered separate from our physical lives, Clark argues.
“A person’s likeness is more than a collection of pixels. Other people use it to recognize them, interpret them, judge them, desire them, trust them, and remember them,” he says. “So, a synthetic image can be technically false and still produce entirely real consequences.”
When someone’s image is depicted in a way they didn’t sanction and can’t control, “That injury goes deeper than deception,” he insists. “There’s real harm being done to the individual.”
Clark, who taught an AI image modeling workshop for New York Tech Week in May, is neither a lawyer nor a legislator. But he believes the people creating these images and videos should be held legally accountable — along with the app’s builders and distributors.
“Our laws already recognize that conduct carried out through a screen can constitute fraud, harassment, theft, exploitation, or abuse,” he reasons. “Nonconsensual sexual deepfakes belong in that same understanding.”
But we’d better regulate it fast, he warns: “This technology is primed to get better and will become harder to distinguish from reality. And when there’s no way to prove that they are deepfakes … what does that mean for the people involved?”
It means we may want to guard our online photos. Before we find ourselves all together. In our altogethers.
