When deepfakes emerged on the Internet, people gasped at this new tool for deception and wondered how they could ever separate fake from real videos. After all, anyone who has seen the deepfake videos of Presidents Obama and Nixon knows that words can now be convincingly put into mouths that never uttered them.
But deepfakes are not undetectable or unprecedented. And, with the right technical evidence, they can be redressed in the legal system under laws that recognize the harm they cause.
Tech companies have already developed various deepfake detection techniques. These methods of course will need to keep up with “advances” in deception techniques, as they have with other fraudulent practices, like counterfeiting.
Deepfakes aren’t unique. They are simply the latest in a history of manipulative photographic techniques that trace back to the origins of photography.
Some of the earliest image manipulation methods were mild by today’s standards, such as fun-house type mirror distortions, or multiple exposures that put the same person in an image multiple times. But some were realistic enough, or so well directed to customers’ emotions that, just like deepfakes, they convinced many viewers to trust the manufactured images.
The seminal image manipulation case of the early days of photography — you could call it the first deepfake case — involved criminal charges brought against photographer William Mumler, a contemporary of Matthew Brady, the early American photographer known for capturing grim post-battle scenes from the Civil War. Mumler, using the rudimentary photography tools of his day, produced and sold portraits which included in the background a cloudy image of one of the portrait subject’s departed loved ones. These photos became known as “spirit photos.”
Initially Mumler prospered, perhaps because his spirit photos gave his customers, many of whom were believers in spiritualism, what they wanted. His business tapered off in his native Boston after an investigator uncovered a pre-existing photo that Mumler had used for some of his “spirit” images. But he later moved to New York, and, in years following the Civil War, again found customers eager for the comfort and consolation of his spirit photos. (The facts in this post are drawn from Peter Manseau’s 2017 book, “The Apparitionists: a tale of phantoms, fraud, photography and the man who captured Lincoln’s ghost.”)
Authorities in New York charged Mumler with fraud in 1869, and a celebrated trial (apparently a preliminary hearing) ensued. One of the prosecutors was Elbridge Gerry (grandson and namesake of a signer of the Declaration of Independence), whose later justice reform activities included replacing the gallows with the electric chair in New York.
The prosecutors appeared overconfident in the case they presented. Most importantly, they proved that photographs could be manipulated to create fake spirit images in the background, but they didn’t prove how Mumler actually did it.
And on their theory of fraud and false pretenses, Gerry tried his best, arguing that the law “grants no license to the cunning man to deceive the simple by artifices, which he proportions to the mental strength of those with whom he had to deal just as the prisoner proportions his drugs to the bodily strength of his victims.” But the deception theory didn’t fit. After all, Mumler’s customers wanted spirit photos, and that’s what he gave them.
Those evidentiary gaps proved fatal to the case. At the end of the trial, Judge Joseph Dowling, though personally feeling “that trick and deception has been practiced by the prisoner,” dismissed the case for inadequate proof.
Modern deepfake prosecutors should take heed of the mistakes in the Mumler prosecution.
Now, as then, it’s not sufficient to prove that a fraud could have occurred; evidence will be needed to show that a particular defendant in a particular case used a particular fraudulent technique. That’s one of the reasons why we can expect video producers to embed codes and digital watermarks in their productions, to help fraud detection. And parties aggrieved by deepfakes will need to employ sophisticated forensic investigators to track down details on the frauds.
Similarly, prosecutors and plaintiffs will need to show actual harm stemming from modern deepfakes. Privacy torts like false light and harassment, the computer hacking laws enacted in the 1980s, and crimes like extortion may provide a sufficient basis for finding cognizable legal harm. If these tools prove insufficient, legislatures may need to enact new laws specifically outlawing deceptive image manipulation.
As to Mumler, after the trial, he went back to his spirit photography studio, although he also branched out into more useful endeavors, including creating the first viable technique for photoengraving, enabling photos to be printed in newspapers.
In 1872, a woman calling herself Mrs. Lindall called on Mumler and asked him to take her portrait, hoping that the spirit of her late husband would appear in the background. Mumler posed her in his studio, took the photo, and developed it. And yes, it came out with a cloudy image in the background of her late husband, resting his hands on her shoulders.
Mrs. Lindall’s real name was Mary Todd Lincoln. That 149-year-old photo of the former first lady flanked by the shadowy image of a certain tall bearded gentleman, is Mumler’s most famous spirit photo. It’s the granddaddy of today’s deepfakes.
Mark Sableman is a partner in Thompson Coburn’s Intellectual Property group. He is the editorial director of Internet Law Twists & Turns. You can find Mark on Twitter, and reach him at (314) 552-6103 or firstname.lastname@example.org.