What’s a “Deepfake”?

by Michael

In recent years the ability to produce video mimicking another person’s voice, face and movement has improved with lightning speed and democratized at a similar rate. The vast majority of the internet-connected world can now produce what’s called a “deepfake”, which is images or videos created using artificial intelligence to resemble the identity of a nonconsenting individual (Definition mine).

In other words, a deepfake is a fake video that looks like someone doing something they didn’t do, typically made without that person’s consent. It’s important to distinguish deepfakes from lesser technologies like photoshop or simple CGI imagery: while anyone can photoshop an image to resemble someone else, and CGI has been used in the past to modify the appearance of an actor in many movies. Deepfakes are distinguishable from these technologies because unlike photoshop and CGI, deepfakes are very convincing to the uninitiated at first glance.

Deepfakes have had some useful functions, like replacing deceased actors in movies, creating funny videos of celebrities doing things they otherwise wouldn’t do, or recreating memories of a loved one after they’ve passed away. However, the potential harm in deepfakes may outweigh the good. It’s not hard to imagine a rogue employee at a news station creating a deepfake of a politician announcing an apocalyptic event and airing it over the network to cause chaos, or for a rejected heir to create a “video will” of their deceased loved one where the deceased renounces their will and leaves their entire estate to the formerly-rejected heir despite the actual wishes of the deceased. Some younger content creators have even found deepfakes of themselves performing sexual acts despite the family-friendly nature of their content.

The purpose of a deepfake is often to fool people, which is why states have begun criminalizing deepfakes of nonconsenting individuals. It’s important to note that already the states don’t seem to agree on their deepfake jurisprudence; between the five states that have criminalized deepfakes (not including election-related crimes), two classify it as a form of sex crime, one classifies it as a privacy crime, one classifies it as an obscenity crime and one classifies it as an identity theft crime. Deepfakes are so versatile in their ability to cause harm that it’s unclear how jurisprudence around this technology will evolve until the first civil and criminal cases are filed.