AI Deepfakes

The Goose Is Cooked?


This blog post is edited by Joseph Dunklin, Madilyn Kuperman, and Yongli Yang.

Introduction to Deepfakes

The term ‘deepfake’ is an amalgamation of two words, deep learning and fake, that denotes a type of artificial intelligence technology.1 Deepfake is widely used to fabricate fake but incredibly convincing videos and audio recordings,2 achieved by using facial mapping and artificial intelligence technology to replace a person’s face from an existing video or image with that of an entirely different person.3 This article overviews the history of deepfakes, the current applications of deepfake technology, the legal issues deepfake technology has created, and possible ways to overcome those issues.

Deepfake first attracted public attention in 2017 when a Reddit user named ‘deepfakes’ circulated pornographic videos that featured celebrity faces swapped onto the faces of the pornographic actors.4 After this incident, a plethora of non-pornographic videos of politicians and actors started to circulate on the internet, including an April 2018 video created by comedian Jordan Peele which purported a deepfake likeness of former President Barack Obama insulting President Donald Trump in a speech.5

Deepfakes can be sourced from as little information as a single still image of the subject, as shown by the work of researchers at Samsung’s A.I. lab in Moscow, who recently used to technology to create clips of Marilyn Monroe, Salvador Dalí and the Mona Lisa.6 And researchers from UC Berkeley have demonstrated that deepfake technology can be expanded to portray movements of the entire body, even to the extent required to mimic the work of professional dancers.7 A Japanese company, DataGrid, has created high-resolution whole-body images of non-existent people using similar technologies8

At this time, many of the deepfakes available on the internet are pornographic in nature–research by Deeptrace found that 96% of deepfakes were non-consensual pornography–often with a computer-generated face of a celebrity replacing that of the original adult actor in a scene of sexual activity.9

Issues Created by Deepfakes

Malicious use of deepfakes can lead to serious consequences. For example, fake videos showing targeted politicians slurring racial comments or taking bribes can be created to sabotage elections.10 To make things worse, making deepfake videos of politicians is relatively easy, since there are a lot of photos and videos of politicians on the internet that could be used as sources.11 The corporate world has also expressed concerns that deepfakes showing false statements of material events such as mergers or unexpected financial losses may be used for market and stock manipulation.12

Individuals can also be victims of deepfakes and it has already become reality – in 2018, an Australian high-school student found out that someone inserted her face in pornographic videos and photos after she googled herself.13

In legal settings, deepfakes can pose a severe threat to the authenticity of crucial evidence in a courtroom, because the continuous advancement of technology makes it significantly more difficult for the average person to differentiate between a real video and a deepfake.14 Some courts have allowed photographic evidence to be admitted without verification of accuracy from eyewitnesses under the “Silent Witness Theory.”15 If the circulation of deepfakes becomes commonplace, lawyers could claim that a video or recording against their client is a deepfake, which may lead juries to question the authenticity of perfectly legitimate video evidence. There are already cases in which a suspect accused of possessing child pornography has claimed that it is computer-generated.16 If this trend continues, it will become increasingly difficult to rely upon key videos–smartphone recordings of police brutality, for example–in the courtroom.

Deepfakes and Laws

A flat-out ban on deepfakes seems to be the most effective way to eliminate all the problems mentioned above, but it would violate the conception of freedom of expression as a fundamental individual right. This Right is documented as a human right under article 19 of the Universal Declaration of Human Rights (UDHR)17 and International Covenant on Civil and Political Rights (ICCPR).18 Therefore, deepfakes have challenged legislative systems across the globe.

In the U.S., injunctions against deepfakes are likely to face First Amendment challenges.19 Even if such injunctions survive a First Amendment challenge, lack of jurisdiction over extraterritorial creators of deepfakes would inhibit their effectiveness.20Therefore, injunctions against deepfakes may only be granted under few specific circumstances, including obscenity and copyright infringement.21

Legislation that aims at issues involving deepfakes may be another effective solution. Three states in the United States have already enacted laws that focus on some issues caused by deepfakes. In 2019, Texas became the first state to ban political deepfakes.22 Virginia followed suit later that year, banning deepfake pornography by making amendments to criminalize revenge porn.23 California also makes it illegal “to create or distribute videos, images, or audio of politicians doctored to resemble real footage within 60 days of an election.”24 This law aims at protecting voters from misinformation, but one scholar remains skeptical about whether it will be enforced.25

At the federal level, a bill known as the DEEP FAKES Accountability Act is under consideration.26 DEEP FAKES is an acronym for “Defending each Person from False Appearances by Keeping Exploitation Subject.”27 If passed, this law would make it a crime to create a deepfake without including a digital watermark or text description indicating that it has been modified.28 A civil penalty of up to $150,000 would also be imposed for each purposeful failure to provide such watermark or description.29

The most feasible short-term solution may come from tech giants like Facebook, Google and Twitter, who can take action to limit the spread of harmful deepfakes. For example, Facebook has already announced that it will remove deepfakes and other manipulated videos from its platform.30

In Europe, specific laws targeting deepfakes are generally lacking, but deepfakes may be limited by laws that target disinformation. In Germany, the Network Enforcement Act, which was enacted in 2017 to fight dissemination of false information, may be invoked to ban deepfakes if it is used to circulate fake news,31 and Spain32 and France33 have enacted similar laws. The European Commission, when discussing strategies to tackle online disinformation, pointed out that deepfakes have been used to create disinformation and manipulate public opinion,34 and it has provided a guideline on how to tackle online disinformation, including requiring online platforms to adopt measures to protect users from disinformation.35 Creation of deepfakes could also violate other European laws, exemplified by an incident in UK where a city worker was convicted of harassment when he created a gallery of fake pornographic images of a colleague.36

In China, the Cyberspace Administration has released rules which took effect on January 1, 2020 that ban any online publication of false information that lacks disclosure that the media has been altered using Artificial Intelligence or Virtual Reality techniques.37

Similar to Europe, India does not have any specific laws that target deepfakes.38 Relevant issues can be covered under sections 67 of The Information Technology Act 2000, however, that provides punishment for publishing sexually explicit content in electronic form.39 Section 500 of the Indian Penal Code 1860, which criminalizes defamation, may also provide limited punishment against deepfake creators.40 Additionally, the Personal Data Protection Bill 2019, if passed, may provide additional protections for individuals, since the proposed legislation imposes restrictions on how personal data can be used and processed.41

Conclusion And The Way Forward

Deepfakes are a global phenomenon, and a solution to the problems they can create is far from clear. What is clear is that we need to take some rigorous steps to address issues created by them. As discussed in this article, although existing laws may restrict circulation of deepfakes under certain circumstances, specific legislation that aims at limiting the potential damage of deepfakes is highly necessary.

On the other hand, governments should focus on increasing public awareness of relevant issues, because legislation cannot prevent all deepfakes from circulating, and legal remedies, no matter how good and efficient they might be, can only be applied after the damage has already been done.

a 2nd Year B.A.L.L.B. (H ons.) Candidate, Dr. Ram Manohar Lohia National Law University, Lucknow, India.

1. Deepfakes Are the Evolution of Fake News and Are Equally as Dangerous, Ste Davies, (last visited July 9, 2020).

2. Ian Sample, What Are Deepfakes – And How Can You Spot Them?, Guardian (Jan 13, 2020,
5:00 AM),

3. Mika Westerlund, The Emergence of Deepfake Technology: A Review, 9 Tech. Innovation Mgmt. Rev. 39, 39, 40 (2019),

4. Rise of the Deepfakes, Week (June 9, 2018),

5. James Vincent, Watch Jordan Peele Use AI to Make Barack Obama Deliver a PSA about Fake News, Verge (Apr. 17, 2018, 1:14 PM),

6. Simon Parkin, ‘Politicians Fear This Like Fire’: The Rise of the Deepfake and the Threat to Democracy, Guardian (June 22, 2019, 8:00 AM),

7. James Vincent, Deepfakes for Dancing: You Can Now Use Ai to Fake Those Dance Moves You Always Wanted, Verge (Aug. 26, 2018, 2:00 PM),

8. Chris Burt, DataGrid Develops Ai to Generate Whole-Body Images of Nonexistent People, Biometric Update (May 3, 2019),

9. Giorgio Patrini, Mapping the Deepfake Landscape, Deeptrace (July 10, 2019),

10. Westerlund, supra note 3, at 42.

11. Id. at 43.

12. Id.

13. Ally Foster, Picture Reveals Sickening Online Secret, (JUNE 30, 2018, 7:33AM),

14. Theodore F. Claypoole, AI and Evidence: Let’s Start to Worry, Nat’l L. Rev. (Nov. 14, 2019), https://

15. See id.

16. Kaveh Waddell, The Deepfake Threat to Evidence, Axios (Oct. 12, 2020),

17. See G.A. Res. 217 (III) A, Universal Declaration of Human Rights (Dec. 10, 1948). https://

18. See G.A. Res. 2200 (XXI) A, International Covenant on Civil and Political Rights (Dec. 16, 1966),

19. Jessica Ice, Defamatory Political Deepfakes and the First Amendment, 70 Case W. Res. L. Rev. 417, 418 (2019).

20. Id.

21. See Id. at 440–41.

22. Allie Morris, Texas Is First State to Ban Political ‘Deepfake’ Videos, San Antonio Express News (Oct. 9, 2019),

23. Virginia Bans ‘Deepfakes‘ and ‘Deepnudes‘ Pornography, BBC News (July 2, 2019), https://,videos%20

24. Kari Paul, California Makes ‘Deepfake’ Videos Illegal, But Law May Be Hard to Enforce, Guardian (Oct. 7, 2019, 18:42),

25. Id.

26. See Defending Each and Every Person from False Appearances by Keeping Exploitation Subject to Accountability Act of 2019, H.R.3230 – 116th Cong. (2019).

27. See id.

28. Mathew Ingram, Legislation Aimed at Stopping Deepfakes Is a Bad Idea, Colum. Journalism Rev. (July 1, 2019),

29. David Ruiz, Deepfakes Laws and Proposals Flood USMalwarebytes Labs (Jan. 23, 2020),

30. Nandita Bose, U.S. Lawmakers Say Facebook Steps to Tackle Deepfake Videos Not Adequate, Reuters (Jan. 8, 2020, 10:35 AM),

31. Initiatives to Counter Fake News: Germany, Libr. Cong.,
germany.php (last visited July 29, 2020).

32. Tim Cushing, Spanish Government Moves Ahead with First Fake News Prosecution, Techdirt. (Mar. 3, 2020),

33. Michael-Ross Fiorentino, France Passes Controversial ‘Fake News’ Law Euronews (Nov. 22, 2018),

34. Tackling online disinformation: a European Approach, at 5, COM (2018) 236 final (Apr. 26, 2018),

35. See id. at 7.

36. Mei-Ling Huang, Deepfakes and the Law – What Can You Do If Your Face Appears in a Deepfake Video?, Royds Withy King (June 18, 2019),

37. Nick Statt, China Makes It a Criminal Offense to Publish Deepfakes or Fake News Without Disclosure, Verge (Nov. 29, 2019),

38. Simran Jain & Piyush Jha, Deepfakes in India: Regulation and Privacy, S. Asia @ LSE (May 21, 2020),

39. Id.

40. Id.

41. Id.