Abstract
Deepfakes—a combination of the words “deep learning” and “fakes”—is a twenty first century term for images, video, and audio recreations of the image and likeness, and sometimes the voice or performance attributes, of celebrities, politicians, and other persons. Contemporary generative artificial intelligence (AI) tools for image creation (e.g., DALL-E 3, Midjourney, Stable Diffusion), AI generated video (E.g., Open AI’s Sora, VideoGen, RunwayML), and AI generated audio, voice, and musical performance (e.g., Suno, RVC WebUI, Udio, Altered, ElevenLabs), have increased the speed and ease with which people can “fake” the appearance, voice, performances, and actions of real people. Whether one views these generative AI tools as a massive step toward the democratization of creation or a ridiculously fast and easy way to exploit the good will of companies and celebrities, interfere with the reporting of facts, or commit a myriad of crimes, the technology is present and only going to improve from where it is today.
Given that deepfakes are almost always associated with expression of some kind, any attempt to litigate or regulate deepfakes will require the consideration of the First Amendment. The Jack Daniel’s v. VIP Products U.S. Supreme Court case reinterpreted or at least clarified the proper scope and application of the Rogers v. Grimaldi test which, since 1989, has been used as a test of First Amendment-based fair uses in federal trademark law infringement, false endorsement, false designation of origin, and trademark dilution claims, and occasionally in state law name-image-likeness exploitation claims under right of publicity or right of privacy theories. The Court’s attention to the Rogers test potentially affects the First Amendment fair use analyses in each of these areas of law. This article will examine the past, present, and future use of the Rogers test as a test and evaluation of the application of First Amendment protections in trademark cases and in legal actions traditionally associated with the exploitation of celebrities, sports figures, and other persons.
Jack Daniel’s held that the Rogers test may not be applied at the motion to dismiss stage to determine if the use is even actionable under the trademark laws so as to save the defendants the cost of developing the infringement case. The test might continue to be applied at the infringement stage, but only if the defendant did not use the plaintiff’s trademark or trade dress in a “source-identifying” way—i.e., as a component of the defendant’s own marks, trade dress, or promotional material. Jack Daniel’s also held that parody and other comment and criticism First Amendment defenses also are diminished if the user made use of another entity’s marks and trade dress in the second user’s own expression. The recent cases of Hermès Int'l v. Rothschild (Metabirkins) and Van’s v. MSCHF (Wavy Baby) are indicative of this situation and serve as predictors of how cases with this type of expressive but potentially confusing or exploitative use will fare, namely that the uses will not be protected by First Amendment parody or comment and criticism uses when the plaintiff’s marks, trade dress, and general good will were commercially exploited by the defendant.
The Supreme Court’s clarifications of when Rogers applies and to what extent the context of a trademark source-identifying use defeats the application of Rogers and limits the application of parody and other free expression defenses are not controlling and have less importance in state law actions for right of publicity and right of privacy claims. But the parallel nature of trademark false endorsement and false designation of origin and state law publicity and privacy actions should mean that a “trademark” type source-identifying use of a person’s name, image, likeness, or other identifying information that is by definition highly commercialized and directly targets the person’s privacy rights or publicity rights for exploitation should fail the balancing tests used in state law claims, of which the Rogers relatedness test is one. These tests would be likely to balance this level of exploitation against a finding of fair use even if the use was embedded in a related expressive activity.
Given that deepfakes are almost always associated with expression of some kind, any attempt to litigate or regulate deepfakes will require the consideration of the First Amendment. The Jack Daniel’s v. VIP Products U.S. Supreme Court case reinterpreted or at least clarified the proper scope and application of the Rogers v. Grimaldi test which, since 1989, has been used as a test of First Amendment-based fair uses in federal trademark law infringement, false endorsement, false designation of origin, and trademark dilution claims, and occasionally in state law name-image-likeness exploitation claims under right of publicity or right of privacy theories. The Court’s attention to the Rogers test potentially affects the First Amendment fair use analyses in each of these areas of law. This article will examine the past, present, and future use of the Rogers test as a test and evaluation of the application of First Amendment protections in trademark cases and in legal actions traditionally associated with the exploitation of celebrities, sports figures, and other persons.
Jack Daniel’s held that the Rogers test may not be applied at the motion to dismiss stage to determine if the use is even actionable under the trademark laws so as to save the defendants the cost of developing the infringement case. The test might continue to be applied at the infringement stage, but only if the defendant did not use the plaintiff’s trademark or trade dress in a “source-identifying” way—i.e., as a component of the defendant’s own marks, trade dress, or promotional material. Jack Daniel’s also held that parody and other comment and criticism First Amendment defenses also are diminished if the user made use of another entity’s marks and trade dress in the second user’s own expression. The recent cases of Hermès Int'l v. Rothschild (Metabirkins) and Van’s v. MSCHF (Wavy Baby) are indicative of this situation and serve as predictors of how cases with this type of expressive but potentially confusing or exploitative use will fare, namely that the uses will not be protected by First Amendment parody or comment and criticism uses when the plaintiff’s marks, trade dress, and general good will were commercially exploited by the defendant.
The Supreme Court’s clarifications of when Rogers applies and to what extent the context of a trademark source-identifying use defeats the application of Rogers and limits the application of parody and other free expression defenses are not controlling and have less importance in state law actions for right of publicity and right of privacy claims. But the parallel nature of trademark false endorsement and false designation of origin and state law publicity and privacy actions should mean that a “trademark” type source-identifying use of a person’s name, image, likeness, or other identifying information that is by definition highly commercialized and directly targets the person’s privacy rights or publicity rights for exploitation should fail the balancing tests used in state law claims, of which the Rogers relatedness test is one. These tests would be likely to balance this level of exploitation against a finding of fair use even if the use was embedded in a related expressive activity.
| Original language | American English |
|---|---|
| Journal | Indiana law review |
| Volume | 58 |
| Issue number | 1 |
| State | Accepted/In press - Jun 10 2024 |
Keywords
- Deepfakes, Generative AI, Image creation, Voice performance, Exploitation, First Amendment, Rogers test, Trademark infringement, False endorsement, False Designation of Origin, Right of publicity, Jack Daniel’s v. VIP Products, Parody, Hermès v. Rothschild, Commercial exploitation, Privacy rights
