Regulating Deep Fake: Use Cases & Future

10. December 2021

Deep Fake & Its Origin

Does anyone still recall the 1997-blockbuster Face/Off? In the movie, Nicholas Cage and John Travolta’s characters go through surgery and end up wearing each other’s faces. Some 20 years later, face-swapping can be easily achieved. Not by face swap surgery, but by “deep learning” and “fake”, aka, deep fake.

Deep fake technology enables swapping the face of a person in an existing image or video with someone else’s and appears incredibly realistic. The pillar of deep fake is deep learning and generative neural network architectures training. For example, one of the first deep fake apps, FakeApp, was built based on Generative Adversarial Networks (GANs) or Autoencoders (AEs) ideas. 

The Misconducts of Deep Fake

Since its birth, deep fake has been linked with great controversy.  Democratization of deep fake makes its usage possible for everyone and publishes their “work” on the internet. The availability triggered harmful uses of the technology. The most well-known is nonconsensual pornography (accounts for 90%-95% of Deep fake videos online), political campaigns based on fake news, financial fraud, etc.…

Such use cases usually do not have the subjects’ consent, which violates their privacy, not to mention the psychological harm on the subjects and the damage to their reputation. When the “protagonist” of the deep fake video is a public figure with influence, the impact of fake messages spoken out of “their mouths” is hard to measure. It should not be overlooked because now more than ever, the public are victims of the “confirmation bias”: people only want to believe in what they already believe and are not willing to think twice to separate the truth from false. Misleading information created by deep fake is not helpful for this tendency. 

Positive Use Cases of Deep Fake

However, not all use cases of deep fake are negative. Here are some positive deep fake use cases and examples:

  • Education: Deep fake can be used for educational purposes. It can make the teaching process more fun and interactive for students. For example, the Illinois Holocaust Museum and Education Center created holographic interviews, enabling visitors’ interaction with Holocaust survivors.
  • Accessibility: Sometimes, language barriers are one of the reasons why important messages cannot be spread on the global level. Deep fake can replicate voices as well as videos. It is possible to create videos in which a person addresses an issue in multiple languages with his/her own voice, even if the person cannot speak some of the languages. This video of David Beckham launching the Malaria Must Die campaign.
  • Training: the problem of data bias is starting to appear with the development of AI and facial recognition technology. It is easier for the algorithm to recognize caucasian males than other ethnic groups. The lack of diversity in training data can be solved by the “virtual humans” created by Deep fake.
  • Creativity: most of us probably have seen an interesting clip of a celebrity face swapped into scenes from a popular movie at some point. These videos are created without sophisticated reason, simply out of fun. However, these videos may be involved in IP-related controversies[1]

Deep Fake and Its Future Legal Approaches

The increasingly severe impact of deep fake misuses urges the legislators to take action against the new technology. However, banning it completely is not realistic, considering its benefits when used for positive purposes[2].  

An approach to misuses of deep fake is to put it into context. Deep faked non-consensual pornography created the most severe impact, and made the most victims. Forty-six states in the US banned revenge porn, but only Virginia and California included faked and Deep faked media[3]

A possible way to charge non-consensual usage of deep fake is to use IP law. However, it can only happen if the victim’s face is derived from a copyrighted photo. If the victim can prove the media maker/publisher intends to harm him/her, harassment law can be used. However, according to Mania, it is often impossible to gather such evidence[4].  

Another approach is to take a market-driven solution[5]. It is suggested that companies who run content dissemination platforms develop algorithms to detect deep fake technology, and create a watermarking system. If such a system is applied, it is easier to categorize the maker/publisher’s liability when using Deep fake[6]

However, a comprehensive regulation towards deep fake is still in the starting phase, despite the widely-acknowledged necessity of introducing such legislation.

Deep Fake & Data Privacy

There are some use cases of protecting subjects’ privacy using deep fake. For example, deepfaking images and videos of journalists and human rights activities to keep them from danger and persecution. However, the most concerning issue of deep fake is non-consensual usage. It violates the subject’s privacy and is a cybersecurity threat. Governments of different countries are starting to respond to deep fake induced civil and criminal challenges, but there has not been much legislative change yet. The democratization of AI would make deep fake available to the vast majority of the public, urging for legislation to regulate its usage and protect individuals’ privacy.

A method to protect your personal information online is data anonymization. Our Deep Natural Anonymization solution based on synthetic AI is able to protect identities by creating synthetic face overlays to protect individuals from recognition, but at the same time keep data quality for machine learning. Data anonymization is an effective method to protect individual privacy, and is compliant with data protection regulations.

If you want to learn more about data privacy and how to protect your personal data using anonymization, please contact us here.

 

[1] Meskys, Kalpokiene, Jurcys, Liaudanskas; Journal of Intellectual Property Law & Practice; “Regulating deep fakes: legal and ethical considerations”;2019

[2] Boggs; Lexology; “The Rise of the ‘Deepfake’ Demands Urgent Legal Reform in the UK”; 2021-03-23

[3] Hao; MIT Tech Review; “Deepfake porn is ruining women’s lives. Now the law may finally ban it.”; 2021-02-12

[4] Hao; MIT Tech Review; “Deepfake porn is ruining women’s lives. Now the law may finally ban it.”; 2021-02-12

[5] Meskys, Kalpokiene, Jurcys, Liaudanskas; Journal of Intellectual Property Law & Practice; “Regulating deep fakes: legal and ethical considerations”;2019

[6] Schmidt; IAPP; “Privacy law and resolving ‘deepfakes’ online”; 2019-01-30

Xinzhuo Xiao
Marketing & Communication
xinzhuo.xiao@brighter.ai