Among the various types of AI-generated works that are being created and marketed nowadays are works that replicate the sound or visual images of specific artists. In many cases, these works, or the models or tools that helped generate them, are marketed using the original artists’ names. To the extent that trademarks subsist in those names, the Supreme Court’s recent decision in Jack Daniel’s Properties, Inc. v. VIP Products LLC could strengthen claims for trademark infringement. This article examines what Jack Daniel’s — and in particular the Supreme Court’s approach to infringement analysis in the context of arguably expressive or artistic uses of other parties’ trademarks — might mean for these use cases.

Trademark Usage in “Sound-Alike” AI Projects

While AI models and tools will undoubtedly yield huge swaths of new and original content, a current trend in the generative AI space is to create works that replicate or mimic the sound or image of a celebrity, artist or music group, and then to distribute the work using the name of that individual or group. Earlier this spring, for example, the AI-generated song “Heart on My Sleeve” generated tremendous media attention — and more than a little concern about the broader impact of AI — when it was released as AI simulations of popular musicians Drake and The Weeknd.¹ Although frequently acknowledged as being AI-generated, the song also was often described or titled using Drake’s and The Weeknd’s names, with those names explicitly leveraged to increase the song’s visibility.²

• In some cases, the creators of these AI tools, models and works are simply hobbyists experimenting with this new technology and distributing the resulting works to highlight their own talents or what the technology can offer.
• In other cases, model and tool builders and end-user creators are exploiting the names of the original artist for financial gain, such as through the sale of advertising or subscription revenues. For example, certain platforms and websites hosting subscription-based or pay-based AI models generate traffic and revenues by offering users the ability to simulate the voices, lyrics and styles of particular artists, offering a smorgasbord of famous artists or musicians from which to choose. In those instances, it is common to see artists’ names and images used to promote the models or tools.

Trademark Law and the Jack Daniel’s Decision

The core purpose of U.S. trademark law is consumer protection. While trademark rights give brand owners a degree of control over names, images and symbols that denote those owners’ goods and services, liability for trademark infringement arises only when another party’s use of those names, images or symbols creates a likelihood of consumer confusion about the source, approval or affiliation of that other party’s goods or services. Federal courts typically analyze likelihood of confusion by considering an array of potential factors, such as the strength of a plaintiff’s trademark, the proximity of the plaintiff’s and defendant’s goods and their respective marketing channels, the defendant’s intent and the sophistication of the relevant consumers.

When addressing the use of trademarks in traditionally “expressive” and artistic works (e.g., books and movies), some courts recognize more pressing First Amendment concerns and thus have applied a heightened standard whereby a plaintiff must demonstrate either (i) the defendant’s use of the trademark is not artistically relevant (in which case the standard multi-factor confusion analysis will apply), or (ii) it’s “explicitly misleading” to consumers as to the source or affiliation of the defendant’s work.

The foregoing analysis — frequently referred to as the “Rogers test” given its origins in the Second Circuit’s 1989 decision in Rogers v. Grimaldi³ — has been a useful tool for putative trademark infringement defendants, particularly in jurisdictions that have taken a fairly expansive view of the scope and application of that test to effectively immunize certain trademark uses from potential liability.

The Supreme Court’s unanimous decision in Jack Daniel’s, however, casts significant doubt on the viability of a defense based on the Rogers test.

Jack Daniel’s concerned a humorous “Bad Spaniels” dog toy that imitated the shape of the Jack Daniel’s bottle. In 2020, the Ninth Circuit Court of Appeals reversed a bench trial verdict finding that the dog toy infringed and diluted Jack Daniel’s trademarks, reasoning that the dog toy was “expressive” and thus the lower court erred by not applying the Rogers test. The Ninth Circuit’s extension of the Rogers test to the defendant’s consumer product reflected a departure from courts that had historically limited the test’s application to more traditional expressive works, and extended it to use of the mark on commercial goods where the mark was arguably used as a source of origin.

The Supreme Court rejected the Ninth Circuit’s position. While disclaiming any ruling on the merits of the Rogers test as a general matter, the Court noted that the test “has always been a cabined doctrine,” and a defendant’s mere “expressive” use of a trademark — for example, conveying humor or parody — does not automatically lead to application of the test. Rather, and regardless of any expressive purposes, the Rogers test has no application where an alleged infringer uses a trademark “as a designation of source for an infringer’s own goods.” In fact, the Rogers test only “kicks in when a suit involves solely nontrademark uses of [a] mark—that is, where the trademark is not being used to indicate the source or origin of a product, but only to convey a different kind of message.” If a defendant “may be trading on the good will of the trademark owner to market its own goods,” the Rogers test “has no proper role.”

Jack Daniel’s Relevance to AI-Generated Works

Notwithstanding the focus on copyright law and rights of publicity when addressing potential causes of action against creators and distributors of AI-generated works, trademark infringement and dilution claims may be options for putative plaintiffs as well. To the extent that an artist’s name also serves as a trademark that identifies the source of that artist’s work (or other goods or services), whether federally registered or rooted in common law rights, a trademark-based, false endorsement or unfair competition may lie where the AI tool or AI-generated work creates consumer confusion about source, approval or affiliation. When facing such a claim, one could easily imagine a defendant raising First Amendment defenses, including that the difficult-to-satisfy Rogers test should apply because the AI-generated work is expressive and artistic.

That defense has potentially been weakened by the Supreme Court’s repudiation of an automatic application of the Rogers test for expressive works. Plaintiffs in AI-generated work cases may, depending on the circumstances, be in a position to fairly argue that the use of the artist’s name and/or trademarks in connection with the AI-generated work serves as a designation of source, thus foreclosing the application of the Rogers test at all. Certain plaintiffs may have a reasonable basis to contend that the party that creates or distributes the AI-generated content is “trading on the good will of the trademark owner” to increase the popularity and commercial attractiveness of the AI-generated work or model (or a platform that hosts those materials). To use the example above, the notoriety of “Heart On My Sleeve” has sprung from the fact that the song is identifiable — and is expressly identified — as AI versions of Drake and The Weeknd. The same would apply to models or tools that are fine-tuned to a specific artist and promoted with that artist’s name and likeness.

The viability of a trademark infringement claim necessarily depends on the particular facts, including the extent to which a plaintiff can establish that they actually own trademark rights in their name or other symbols, and ultimately demonstrate a likelihood of confusion. Moreover, a putative trademark plaintiff also may be presented with other defenses in the context of AI-generated works, the viability of which will need to be determined by courts tackling this new issue in the coming years. But it is fair to conclude that the Supreme Court’s ruling in Jack Daniel’s has narrowed one path that defendants in this context may have considered. Creators and distributors of AI models and AI-generated content cannot bank on a court concluding that the Rogers test applies — and potentially using that as a basis to grant a motion to dismiss prior to extensive (and expensive) discovery in an infringement proceeding — merely because those models and works reflect artistic expression.

ENDNOTES

1 https://www.nytimes.com/2023/04/19/arts/music/ai-drake-the-weeknd-fake.html.

2 For example, one YouTube post is titled “Drake – Heart On My Sleeve (feat. The Weeknd),” with the use of AI only revealed in the subsequent description text below the video. See https://www.youtube.com/watch?v=pIJSKxVJppA.

3 Rogers v. Grimaldi, 875 F.2d 994 (2d Cir. 1989).

This post comes to us from Skadden, Arps, Slate, Meagher & Flom LLP. It is based on the firm’s memorandum, “How the Supreme Court’s Decision in Jack Daniel’s May Impact Certain AI-Generated Works,” dated June 15, 2023, and available here.