Digital Bodies,
Real Harm

The Northwestern article on deepfake privacy does something rare in legal scholarship. It does not just theorize harm, but it goes out and asks people what they actually think. Across three large survey studies, the authors show that ordinary people see deepfake pornography as profoundly harmful, morally blameworthy, and deserving of serious punishment. They also show that people are often ready to punish even some labeled and consensual deepfakes, and that they treat deepfake pornography as roughly equivalent to traditional nonconsensual pornography. That moral instinct is out ahead of our doctrine. Current privacy torts, defamation law, and even most nonconsensual pornography statutes simply do not reach many of the harms the participants instinctively recognize. This essay argues that the public’s attitudes, as reported in “deepfake Privacy,” should push courts and legislators toward a new framework for sexual privacy, one that treats deepfake pornography as a serious dignitary and autonomy violation that can justify both civil and criminal regulations even in the shadow of the First Amendment.1

I. What The Studies Actually Show About Deepfakes

The authors conducted multiple survey experiments with nationally balanced samples recruited through professional panels. In the main study, 1,141 participants were randomly assigned to four basic conditions: pornographic or attitudinal deepfakes, each either labeled as fake or unlabeled.2 Within each condition, respondents read a set of short vignettes about Will, who uses an app to paste faces onto videos, and then rated each scenario on moral

blameworthiness, harm to the subject, and appropriate punishment. They also answered broader policy questions at the end of the survey. The pornographic scenarios were brutal in their clarity. In the default case, Will takes public photos of his friend Jane, pastes her face onto a naked woman having sex, and posts the video with her entire name. The video looks real and Will does not say it is fake. Participants rated this conduct as very blameworthy and very harmful, with the average scores above 5 or 6 point scale, and they overwhelmingly wanted criminal punishment. Only around seven or eight percent thought this should not be a crime. Even written pornographic stories about Jane were seen as quite serious, though slightly less harmful and less deserving of punishment than the video.3

_________________________

1.​ Matthew B. Kugler and Lyrissa Barnett Lidsky, “Deepfake Privacy,” Northwestern University Law Review 116 (2021): 611.​ 2.​ Ibid., 636–38.​ 3.​ Ibid., 638–40.

The authors then varied the facts. Sometimes the deepfake subject was a celebrity like the header photo of Taylor Swift rather than a friend. Sometimes there was no nudity, only sexualized behavior like spanking or seductive language. Sometimes the video was created only for personal use rather than public posting. Surprisingly, these variations barely mattered. Whether the subject was a celebrity, whether there was nudity, or whether the video involved BDSM style spanking rather than intercourse, respondents still saw conduct as highly harmful

and criminal.4 The only substantial drop came when the deepfake subject expressly consented to a pornographic deepfake for personal use that was never shared. Even then, the sample split. Many participants still wanted to punish Will, but a large share now said this should not be a crime and that his conduct was not morally wrong. The distribution became bimodal. Some people saw consensual deepfakes as a private sexual choice. Others still saw them as disturbing.5

___________________________

5. Ibid., 641–42.

Labeling told a completely different story. For pornographic deepfakes, adding a clear label like “This is fake” barely moved the needle. There were small reductions in recommended punishment, but no meaningful shift in perceived harm or blameworthiness. For written stories, labeling mattered more. For videos that showed real looking bodies and faces, the label could not neutralize the gut sense that something deeply wrong had been done.6 The attitudinal scenarios were mixed as well. Here the deepfakes made people look like they used cocaine, shook hands with a child molester, endorsed terrorism, insulted themselves, or sang a silly campaign song. None of these videos were sexual in nature but participants still tended to see them as harmful and criminal, especially the cocaine and terrorism cases, but the ratings were lower than pornography and more sensitive to labeling and consent. Labeled attitudinal deepfakes and playful or seemingly harmless ones, like scientist describing their life, were treated as less serious, though most people still supported some form of legal sanctions.7

Two follow up studies sharpened the policy implications. In one, the authors let participants choose between no punishment, a civil lawsuit, criminal punishment, or both. Across several scenarios, most respondents chose both civil and criminal remedies. A smaller, but nontrivial group preferred civil suits only. Even fewer wanted no legal response at all.8 In the final study, the authors directly compared deepfake pornography of an ex partner to traditional nonconsensual pornography where an ex posted a real masturbation video that had been shared in confidence. Participants saw both as extremely harmful and morally wrong. If anything, they rated the deepfake scenario as slightly more blameworthy and slightly more deserving of criminal punishment.9 Taken together, the data show three big things. First, people see pornographic deepfakes as intensely harmful, roughly on the same moral level as traditional nonconsensual pornography. Second, they do not believe that labeling pornographic deepfakes as fake meaningfully cures that harm. Third, they are generally open to both criminal and civil remedies, especially for nonconsensual pornographic cases.

____________________

6. ​Ibid., 639–40, 643.​

7. Ibid., 644–49.​

8. Ibid., 655–57.​

9. Ibid., 658–59.

II. Why People Condemn Even Consensual or Labeled Deepfakes

Some of the findings look puzzling at first. If harm in pornography is about lack of consent, why do so many participants still want to punish consensual deepfakes that are used only for private purposes? And if defamation and false light are built on falsity and reputation, why are labeled deepfakes still seen as so harmful even when viewers are told the video is fake? The article offers two overlapping explanations. One comes from moral psychology. The other comes from feminist and privacy critiques of consent. For moral psychology , the authors draw on research about moral dumbfounding and disgust. People often have an immediate intuitive reaction that something is wrong and only later search for rational justifications.10 Deepfakes are strange, technologically novel, and often used in situations that feel humiliating. That weirdness and disgust can bleed across contexts. Once people have seen a pornographic deepfake in their mind, they may carry that same feeling into much softer, even consensual, cases and still want to condemn them. Work on “dyadic morality” also suggests that once people see a practice as harmful, they will also see it as immoral, and that this harm and wrongness feed each other in a loop.11 Deepfakes look harmful, so they feel wrong and because they feel wrong, people imagine more harm.

___________________

10. Jonathan Haidt, Fredrik Björklund, and Scott Murphy, “Moral Dumbfounding: When

Intuition Finds No Reason” (unpublished manuscript, 2000), discussed in Kugler and Lidsky, “Deepfake Privacy,” 661–63.​

11. Chelsea Schein and Kurt Gray, “Moralization and Harmification: The Dyadic Loop

Explains How the Innocuous Becomes Harmful and Wrong,” Psychological Inquiry 27 (2016): 62–65.

From feminist legal scholarship, the article surrounding deepfake consent in a world where gender and power already shape who is sexualized, who is targeted, and who bears the risk. Pornographic deepfakes overwhelmingly target women, often young women, and are usually created by men. Even where the scenarios describe “consent,” nothing in the vignette tells us that the woman fully understands the technology, the risks of redistribution, or the long tail of images on the internet. The authors connect this to broader feminist critiques of how consent operates under the conditions of inequality, as well as to Daniel Solov’s work on how privacy “self management” breaks down when people cannot really foresee how their data or their image will be used.12 Seen through that lens, it is not irrational for participants to be skeptical of consent in a deepfake scenario. It might be formal consent, but not fully informed or freely given in a meaningful sense. Participants in the survey do not have the vocabulary of feminist theory at hand, but their instincts line up with it. They seem to be saying: once you build this kind of deepfake file of somebody’s naked body, even with their nominal permission, you are lighting a fuse that cannot be put out if it ever leaks, gets hacked, or is reused.

12. Daniel J. Solove, “Privacy Self Management and the Consent Dilemma,” Harvard

Law Review 126 (2013): 1880–1903.

III. Law Behind the Curve: Tort Limits, the First Amendment, and Path Forward

Once we take the survey results seriously, the legal gap becomes painfully obvious. Traditional privacy torts do not really reach deepfake pornography. Intrusion upon seclusion targets invasive access to private spaces, not the reuse of public images in fabricated videos. Public disclosure of private facts requires that the information disclosed be truthful and private, while deepfakes are fabricated from public images. Defamation and false light require falsity, but if a pornographic deepfake is clearly labeled as fake or posted on a site dedicated to fantasy material, courts may see no actionable “false statement” about the subject’s real life.13 The very fact that labeling does not comfort victims or survey respondents exposes the mismatch between doctrine and dignitary harm. The First Amendment is another barrier. The Supreme Court in United States v. Alvarez held that falsity alone is not enough to strip speech of constitutional protection. The government can punish false speech when it is tied to fraud, defamation, or other legally recognized harms, but it cannot criminalize lies simply because they are lies.14 That means lawmakers cannot just outlaw all deepfakes as false videos. They need a more precise interest. The clearest path, and the one article leans into, is to treat deepfake pornography as part of the same family as nonconsensual pornography and morph it under or with child pornography. Dozens of states now have criminal and civil statutes targeting nonconsensual pornography, often called revenge porn. When those statues have been challenged, state supreme courts in Vermont, Illinois, Minnesota, and Texas have upheld them recognized a compelling or substantial state

interest in protecting sexual privacy, preventing reputational and psychological harm, and reducing coercion and harassment.15 Courts have upheld bans on morphed child pornography that pastes a real child’s face onto an adult body. Even though the underlying image is partly fabricated, the use of an identifiable minor’s likeness justifies

______________________

13. Kugler and Lidsky, “Deepfake Privacy,” 660–61.​

14. United States v. Alvarez, 567 U.S. 709 (2012).​

15. State v. VanBuren, 214 A.3d 791 (Vt. 2019); People v. Austin, 155 N.E.3d 439 (Ill.

2019); State v. Casillas, 952 N.W.2d 629 (Minn. 2020); Ex parte Jones, No. PD 0552 18, 2021 WL 2126172 (Tex. Crim. App. May 26, 2021).

treating it like real child pornography, because it exposed that child to reputational and emotional harm and ties them to sexualized images they never consented to.16 If we take the survey seriously, adults victims of deepfake pornography experience a similar type of harm. Their real faces and names are tied to videos that depict them nude, masturbating, or having sex, in ways that threaten their employment, emotional wellbeing, and social standing. Participants in the study rated deepfake pornography as almost as harmful as, and slightly more blameworthy than, traditional nonconsensual pornography. They also strongly favored allowing both civil suits and criminal prosecution.17 That public judgment gives courts a stronger foundation to say what the doctrine has hinted all along: there is compelling state interest in protecting sexual privacy and in preventing

the nonconsensual sexualization of people’s bodies and identities, even when the underlying images are partially fabricated. On the attitudinal side, Alvarez and general First Amendment principals suggest that broad bans on political or satirical deepfakes will be much harder to sustain, especially when they are clearly labeled. But pornographic deepfakes occupy a different space. They are not just false statements. They are forced sexual performances, imposed on someone’s image without their meaningful permission.

_________________________

16. See, for example, Doe v. Boland, 698 F.3d 877 (6th Cir. 2012); United States v.

Hotaling, 634 F.3d 725 (2d Cir. 2011).​

17. Kugler and Lidsky, “Deepfake Privacy,” 655–59.

A sensible legal framework, then, would do at least three things. First, it would create explicit criminal and civil cases as causes of action for nonconsensual pornographic deepfakes, whether or not they are labeled as fake, focusing on distribution and not just creation. Second, it would adopt robust consent requirements like New York’s deepfake statute, including plain language agreements and clear descriptions of the consent and its uses, while still allowing courts to invalidate coerced or uninformed “consent.”18 Third, it would approach attitudinal deepfakes more cautiously, using tools like labeling requirements, election specific protections near voting dates, and strengthened defamation remedies rather than broad criminal bans.

_______________________

18. N.Y. Civ. Rights Law § 52 c(3) (McKinney 2021).

Conclusion

The Northwestern deepfake study makes clear that ordinary people already understand what the law still struggles to name. When someone takes your face and paste it onto a naked body that is having sex on screen, they have done more than lie about you. They have hijacked your sexual identity, your dignity, and your control over how you appear in public. Labeling that video as “fake” does not fix the harm. In some ways it just underlines that your body and your image can be used without your consent. The survey participants' intuition that pornographic deepfakes belong in the same moral and legal category as nonconsensual pornography should not be dismissed as overreaction or confusion. It is a grounded reaction to a new technology that makes it cheap and easy to turn anyone into sexual content. The law of privacy, defamation, and free speech has not yet caught up to this reality. But the data in “Deepfake Privacy” give lawmakers and courts both a map of public judgment and warning. If we continue to treat deepfake pornography as a quirky problem of falsity or satire, we will miss the core injury : the loss of control over one;s sexual self. A modern framework for

sexual privacy should recognize that harm and be willing, even in a speech protective system, to draw a bright line around the nonconsensual sexualization of real people’s images.

Bibliography

Articles and Academic Sources

Chapman, A. Deepfake Harassment and Feminist Harm Theory. University of Mary Washington Research Journal, 2024. Chesney, Robert, and Danielle Keats Citron. “Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security.” California Law Review 107 (2019). Citron, Danielle Keats. “Sexual Privacy.” Yale Law Journal 128 (2019). Citron, Danielle Keats, and Mary Anne Franks. “Criminalizing Revenge Porn.” Wake Forest Law Review 49 (2014). Haidt, Jonathan, Fredrik Björklund, and Scott Murphy. “Moral Dumbfounding: When Intuition Finds No Reason.” Unpublished manuscript, 2000. Kugler, Matthew B., and Lyrissa Barnett Lidsky. “Deepfake Privacy.” Northwestern University Law Review 116 (2021): 611–662. Schein, Chelsea, and Kurt Gray. “Moralization and Harmification: The Dyadic Loop Explains How the Innocuous Becomes Harmful and Wrong.” Psychological Inquiry 27, no. 1 (2016): 62–65. Solove, Daniel J. “Privacy Self-Management and the Consent Dilemma.” Harvard Law Review 126 (2013): 1880–1903. Books / Casebook

Eskridge, William N., Nan Hunter, and Courtney G. Joslin. Sexuality, Gender, and the Law. 5th ed. St. Paul, MN: West Academic, 2023. News / Journalism Beckett, Lois. “California Students Left Reeling after Deepfake Porn.” The Guardian. October 2023.

Statutes / Primary Law Sources N.Y. Civil Rights Law § 52-c (McKinney 2021). Cases Cited American Booksellers Ass’n v. Hudnut, 771 F.2d 323 (7th Cir. 1985).​ Doe v. Boland, 698 F.3d 877 (6th Cir. 2012).​ Ex parte Jones, No. PD 0552 18, 2021 WL 2126172 (Tex. Crim. App. May 26, 2021).​ Lawrence v. Texas, 539 U.S. 558 (2003).​ People v. Austin, 155 N.E.3d 439 (Ill. 2019).​ State v. Casillas, 952 N.W.2d 629 (Minn. 2020).​ State v. VanBuren, 214 A.3d 791 (Vt. 2019).​ United States v. Alvarez, 567 U.S. 709 (2012).​ United States v. Hotaling, 634 F.3d 725 (2d Cir. 2011).