60 Minutes, the CBS weekly news show, did a segment this past Sunday on the powers of so-called deepfakes. Deepfakes refers to the use of AI to alter how a person looks or sounds on video. It can also be used to make one person appear to be and sound like another.
According to Wikipedia. Deepfakes (a combination of “deep learning” and “fake”) are synthetic media in which a person in an existing image on a video is simply replaced with someone else’s likeness. Deepfakes use machine learning and artificial intelligence to manipulate or generate both visual and audio content.
The technique has been used to make celebrities look as if they are in pornography movies. In one example, 60 minutes showed how a Tom Cruise impersonator’s appearance could be altered to make him look just like and sound just like the real Tom Cruise. This alteration was done by superimposing the face of the real Tom Cruise on the impersonator’s face. Since the impersonator already had the mannerisms down pat, the alteration was pretty easy. Here’s a link to the video, as well as nine other famous deepfakes.
It was quickly pointed out that the AI and software are still developing and are presently expensive. See also a recent NPR article that makes the same point. But the creators of this altering process also pointed out that it won’t be long until these alterations are easy.
There is a danger that witnesses and even lawyers could alter their online appearance
This ability got me thinking about the dangers and possible benefits of deep fakes in litigation. As we continue to do things more and more remotely, there is a danger that witnesses and even lawyers could alter their online appearance to become better and younger (or maybe older) looking. Or, for that matter, present a more credible appearance. Since many cases settle short of trial, deep fakes could be used to alter the appearance of deponents to make them appear more credible and enhance their settlement positioning. There is already a deepfake tool that allows users to change their appearance in real-time during live video streams.
And of course, there is the possibility of outright fraud
And of course, there is the possibility of outright fraud: making someone appear to be someone else on a video, for example. One only need to watch the deep fake Tom Cruise TicTok video to see the risks. And there is a reverse issue: will deepfakes give a person who is really on a video an argument that that’s not really them?
I could see in the future a whole new field of expert testimony being developed to spot deep fakes. (and expert witnesses on the subject might even use the tools to make themselves appear more credible).
Perhaps we need new ethical rules to pinpoint the risks and obligations of lawyers
Of course, one answer is to go back to the old days simply: do everything in person. But for a whole lot of reasons, that horse may be out of the barn. So we may come down to using experts and investigation to spot imposters. We also need to educate the judiciary and public about the dangers and risks: we all need to recognize you can’t always believe what you see. And finally, perhaps we need new ethical rules to pinpoint the risks and obligations of lawyers.
Not trying to be chicken little, but the impersonations on the 60 Minute episode were pretty darn impressive. And scary.