Examining Deepfake Ethics and Dangers

Deepfakes have a reputation for being extremely dangerous and of questionable ethics. While any technology can be abused, it is the usage itself that is unethical, not the technology itself. This applies to deepfakes as well. As long as the usage is ethical, deepfakes can be an ethical technology. However, there are still potential threats from deepfakes. The technology is available publicly – anyone can make a deepfake, even those with unethical intentions. The best way to limit the danger of deepfakes may well be to make sure that everyone knows about them, effectively inoculating the public against trusting a video just because they have watched it.

In this chapter, we will cover the ethics and dangers of deepfakes. We will examine the ethics of deepfake’s origins, create some guidelines to follow in creating ethical deepfakes, discuss the dangers of deepfakes, and cover some of the defenses against deepfakes.

We’ll...

The unethical origin of deepfakes

Deepfakes began with an unfortunately unethical beginning. The first public deepfakes were created by a Reddit user by the name of /u/Deepfakes. These first videos showing off Deepfake’s capabilities involved putting (usually) female actors into pornographic scenes. This non-consensual pornography was radically different from previous fakes in that they were more realistic, easier to create, and full video, capable of poses and expressions that the target actor had never performed in any recorded media.

Non-consensual pornographic deepfakes are not victimless events. This usage is extremely damaging as it harms the actors in profound ways. With non-consensual deepfakes being made, celebrities are losing control of their image and reputation. Unfortunately, there is no way to completely avoid the threat from deepfakes as even after proving a video isn’t actually of them, the deepfake will still be spread and objectify the actor. Some...

Being an ethical deepfaker

Deepfakes have many potential uses, both ethical and unethical. Knowing what is ethical or not is sometimes difficult to figure out. For example, are all pornographic deepfakes unethical? What if everyone involved consents? This can be difficult to answer sometimes, and there is a lot of ambiguity, but there are some guidelines that you can follow and some tests you can run to try to ensure your usage is on the right side of the ethical questions.

Consent

Probably the most important ethical concern is one of consent. Consent is quite simple: is everyone involved willing to participate? This is most obviously an issue when it comes to pornographic deepfakes – if not everyone is a consenting participant, then it’s unquestionably unethical.

However, this becomes less obvious when the video is satirical or an actor is put into a role that they didn’t originally perform. Very few deepfake creators can get explicit permission from...

The dangers of deepfakes

The dangers of deepfakes go beyond just fooling some people. If others act on what they see in deepfakes, there are huge dangers to reputation, politics, and even economics. Imagine if a large number of people were convinced by a deepfake of a declaration of war. There could be pandemonium and huge amounts of damage to international reputation before the truth was revealed. This very nearly happened recently: a deepfake of President Zelenskyy of Ukraine was spread online telling the Ukranian people to lay down arms during the Russian invasion.

Reputation

In some ways, danger to reputation sounds like the most trivial of potential dangers from deepfakes. However, reputation can be incredibly important for all kinds of things. Imagine if a person were to lose their job over a deepfake insulting their boss, or another person getting arrested when a deepfake falsely confessing to a crime is sent to the police. These effects can compound and seriously impair...

Preventing damage from deepfakes

There is no foolproof way to make yourself immune to deepfakes or their effects, but there are activities and actions you can take in order to reduce the dangers. In this section, we’ll examine methods that you can use to prevent as much damage as possible.

Starving the model of data

While deepfakes have been done with a minimum of data, the result is not seamless or high quality. This means that if you can simply prevent the deepfaker from getting enough data, you can prevent a deepfake from being of sufficient quality to fool anyone.

If you’re trying to protect a famous person, this tactic is much harder, since every movie, TV show, interview, or even photoshoot that is available is a source of training data for a model. In fact, it’s almost inevitable that enough data is available to target any famous person as the public nature of their job means that a lot of training data will be publicly available.

As a private...

Summary

The ethics and dangers of deepfakes are probably the most publicized aspects of them. However, they’re things that all deepfake creators should consider, as it’s easy to operate unethically if you don’t consider these questions. Even those who aren’t creating deepfakes may need to think about how to prevent damage from deepfakes or other manipulated media.

As a creator, you must consider the ethics of the content you’re creating. Making sure that you evaluate such things as consent, the personal experience, and the labeling of a deepfake is a good way to clear ethical hurdles and make low-risk deepfakes.

Even as non-creators, people should be wary of the threats of deepfakes. The technology is “out of the bag” and cannot be erased. Even if the technology were to be banned immediately, state-level actors would still be able to wield it in potentially dangerous ways and could ignore any laws or restrictions placed on its...