The intersection of artificial intelligence and federal trademark law

Current and future uses of trademarks in Artificial Intelligence (AI) systems could implicate federal trademark law; however, not all scenarios will result in a violation of the Latham Act, nor will they always be immune from such claims. The intersection of AI and trademarks may include rights holders seeking to select marks using machine-generated output, uses permitted under principles of fair use and the First Amendment, or those that infringe or dilute third party rights. Each of these areas are addressed below.

Selection and protection of marks

Typically, the selection of potential trademarks is done by human discussion, an investigation of market preferences and channels, consideration of the goods and services, and creative development of a new term or design. Trademark attorneys search and analyze possible marks and identify potential pitfalls in adopting a certain mark. Many times, a mark searched has been selected by a company or individual with a strong preference for the possible future development of a brand or service under the mark. Upon learning an exact match or a similar mark is in the same or related field, business decisions are made, taking into account the associated risks of adopting that mark. This entire analysis could be avoided altogether for better or worse via AI tools.

Machine selection and clearance of a mark has potential pitfalls, namely, a failure to uncover similar marks, miscalculation of the weakness or strength of a mark or third party use of similar marks, possible parallel overlap in selection by unrelated third parties, and questions regarding the reliance upon a computer-generated search in court or in defense of a claim for profits or willful infringement. Overall, AI lacks human intuition, reasoning, and experience that a trademark practitioner would rely upon in generating and conveying results. Imagine explaining to the court in an infringement action that you relied entirely upon a machine recommendation to establish that a mark was free to use for a particular good or service. See 15 U.S.C. §1117, Romag Fasteners Inc. v. Fossil Inc., et al., 140 S.Ct. 1492 (2020).

Permissible use: Fair use and the First Amendment

Outside of selecting and protecting marks, trademarks are implicated in AI when marks are used in machine learning, data collection and analysis, and certain machine guided outputs. At any stage of the process, a trademark may be used, yet all such uses do not implicate a litigable legal claim under principles of fair use and the First Amendment.

Fair use of a mark may occur where use is descriptive, noncommercial, news reporting and commentary, parodying, criticizing the owner or the goods and services associated with the mark, or in comparative advertising. Playboy Enterprises, Inc. v. Welles, 279 F.3d 796 (9th Cir. 2002); Toyota Motor Sales, U.S.A., Inc. v. Tabari, 610 F.3d 1171, 1175-76 (9th Cir. 2010).

Recently, the Supreme Court addressed fair use and the First Amendment in Jack Daniel’s Properties, Inc. v. VIP Products, LLC No. 22-148 (U.S. June 8, 2023). Bad Spaniels made a dog chew toy that parodied the commercially distinct Jack Daniels whiskey bottle. As a result, Jack Daniels claimed infringement, while Bad Spaniels claimed that its use was parody, performed an expressive function, and therefore, was a fair use of Jack Daniels’ rights and protected under the First Amendment, citing the Rogers v. Grimaldi, 875 F.2d 994 (2d Cir. 1989). The court disagreed, pointing out that the Bad Spaniels use was not merely expressive because it was for commercial purposes, and remanded the case to decide whether it was infringing.

In view of these protections, permissible trademark use in an AI context will be highly fact-specific. If any such use turns on commercial purposes rather than those deemed to perform some descriptive or expressive function, claims under the Lanham Act may be implicated.

Lanham Act deems trademark confusion as infringement

Under the Lanham Act, trademark infringement may be found where there is an unauthorized use that “is likely to cause confusion, or to cause mistake, or to deceive as to affiliation, …or as to the origin, sponsorship, or approval of…goods [or] services.” 15 U.S.C. §§1114(1), 1125(a). Even though new technological uses may arise with unique factual scenarios, they must be considered in view of existing statutory language and case law.

For example, use of marks in AI systems or in algorithms or code not readily apparent to a user may be infringement if the net result is to confuse, cause mistake, or deceive. These situations could be analogous to caselaw regarding the use of meta tags, keywords, and linking on the Internet.

As use of the internet developed with regard to brands, use of keywords and meta tags to alter web traffic became popular, and programmers used competitor names on websites and in source code to alter web traffic. Notably, using third party brand names as meta tags and keywords was found to be trademark infringement where there was consumer confusion and where use of protected marks went beyond what was reasonably necessary to identify the brand. Brookfield Communications v. West Coast Entertainment, 174 F.3d 1036 (9th Cir. 1999); Promatek Industries, Ltd. v. Equitrac Corp., 300 F.3d 808, 812 (7th Cir. 2002).

Infringement considerations in an AI context may also occur in connection with data analytics and outputs. Following Brookfield, initial interest in confusion in an online advertising context was addressed in Multi Time Machine, Inc. v. Amazon.com, Inc., 804 F.3d 930 (9th Cir. 2015). Multi claimed a violation of the Lanham Act in view of Amazon’s search results for Multi’s military style watches called “MTM SPECIAL OPS.” Although Amazon didn’t sell the Multi watches, it listed competing brands of watches. Under this scenario, Amazon avoided Multi’s claims of infringement because it clearly labeled the name and manufacturer of each competing product and included photographs of the items.

Rights holders may also consider theories of post-sale confusion in AI systems even if an immediate consumer is not confused. This occurs where confusion, mistake, or deception may be shown by non-purchasing, casual observers after the point of sale. See Ferrari S.P.A. Esercizio v. Roberts, 944 F.2d 1235 (6th Cir. 1991).

Federal dilution laws

In instances where no confusion is shown, use of a famous mark in an AI context may result in a dilution of rights. See 15 U.S.C. § 1125(c). A complainant must show dilution by blurring or tarnishment. Blurring may occur where the use weakens the distinctiveness of a mark in the minds of consumers by use of a similar mark, and tarnishment may occur where use of the mark is in connection with inferior products or in an unfavorable way, causing reputational harm.

***

In view of the above, AI applications and any yet to be discovered AI systems or media are constantly evolving, therefore, the best strategy to assess the intersection of AI and trademarks is by application of existing trademark laws, cases, and interpretations of each, preferably by an experienced trademark attorney and not another AI system.