Risks

To understand the magnitude of the effects that deepfakes have on our society, we assessed the technology through a risk framework, with an emphasis on the precautionary principle. The goal is to revive collect responsibility among people to be more aware of manufactured risk so that it doesn't escalate.

Precautionary Principle

"When there is risk of significant health or environmental damage to others or to future generations, and when there is scientific uncertainty as to the nature of that damage or likelihood of risk, then decisions should be made so as to prevent such activities from being conducted unless and until scientific evidence shows the damage will not occur."

- Cass R. Sunstein, 2005

Power

“We must make a choice about life in cyberspace - about whether the values embedded there will be the values we want” - (Lessig 546)

Deepfakes can pose a threat to existing power structures—the decisions to distribute technological benefits is a social choice. Malicious intent could skew existing distributions and encourage the spread of misinformation, which poses an even larger threat to society.

What Groups Face the Greatest Harm?

Women

96% of online deepfakes are pornographic, with a disproportionate focus on women. These markets dehumanize women to be replaceable objects and commodities without their consent. The rise of "revenge porn" further accentuates the degradation of women through extortion and invasions of privacy.

Politicians

Political campaigns have seen deepfake use to create content that generate more publicity. Fake election stories receive significantly more engagement and thus are more likely to be pushed into visibility by social media algorithms. Additionally, politicians' voices are easily accessible and could be used to manufacture statements that could incite conflict.

Celebrities

Deepfake content pose a great threat to the credibility of celebrities' brands and businesses. These could include fake nudes or false endorsements of products. Specifically in Hollywood, artificial intelligence poses a great threat to writers, as the data that these AI models are trained on might be infringing on writers' intellectual property.

The Truth is at Risk

There is the general assumption that photographs and video-recorded evidence are reliable and ubiquitous because there is a high volume of genuine information. Deepfakes directly challenge the validity of digital information.

The Epistemic Threat

Even though most people haven't been to Antarctica, we assume that this is a picture of the South Pole because we trust photos to be reliable.

Deepfakes reduce the quality of information in the digital data pool. If humans cannot distinguish a deepfake image from a real image, then the validity of all digital content will be put into question.

via EDsmart.org

Most Americans are unable to detect deepfake content.

These statistics are alarming, not only because it shows how quickly deepfakes have improved, but also because it indicates that most of the public is unaware of what content is real and what is not.

Not only does the ambiguity of online content impact the credibility of the internet as a whole, but it also affects the relationships that individuals have with their governments. Thus, this brings in the question of who should be liable for the consequences.

Considering the fact that Americans view misinformation as detrimental to the country's well-being, increasing deepfake presence in the media raises an alarming flag to Americans and their relationship with the government.

However, the lack of liability hinders the development of an effective solution.

REGULATIONS