INTRODUCTION

Trademarks have been central to commerce since early times, establishing origin, quality, and trust for a business among consumers. In the analogue world, the preservation of these marks was easier; violations were easily discerned; boundaries were more clearly defined, and participants in the violations of a given mark were limited. The digital economy has been revolutionized by AI as it has made significant progress in a short period of time. It has evolved from being just a theory that can analyze data and offer relevant information and insights. 

In modern times, generative AI systems can generate brand names, logos and slogans, and even audio marks that any human counterfeiter could never match in scale and pace. An algorithmically optimized list of millions of products is displayed on e-commerce platforms every day, with numerous brands taking advantage of this feature. As AI-based picture-making software becomes more widely used, there have been growing issues of copyright and trademark infringement that have come up because of the proliferation of these applications. There have been many cases brought before the courts in which an AI-generated piece bears an uncanny resemblance to original intellectual property such as copyrighted materials and trade dress that is protected by trademarks.  

This article addresses three main aspects of the problem:  

  1. The role of AI in enabling and exacerbating trademark infringement;  
  2. The unique danger of brand dilution in AI-fueled environments; 
  3. The new and complex legal and practical hurdles for brand owners and trademark authorities in blocking AI-driven infringement.  

THE GENERATIVE AI PROBLEM

The fundamental question arises that the basis of traditional infringement is seemingly straightforward: are consumers likely to be confused about the source, affiliations, or sponsorship of products or services? However, the issue becomes much more sensitive where infringement results from output created by a machine learning program trained using billions of pieces of data mined from the internet, rather than from the actions of a human infringer.

Generative AI and Trademark Law: Challenges of AI-Generated Brands, Data, and Images

The use of AI technologies increases the likelihood of committing trademark infringement because AI is capable of producing brand-like materials such as logos, trademarks, and images without the users’ knowledge. Businesses could be liable for trademark infringement if the branding confuses consumers and damages the reputation of a well-known brand. The problem is not just theoretical: if an individual asks an AI generator to design a logo for a health and fitness company, the program is going to use the existing training data that likely heavily features existing trademarked logos, and the output might seem confusingly similar to protected logos, even though neither the user nor the AI developer aimed for infringement. 

 Infringement through AI faced by many industries is highlighted in some of the biggest cases.  Hermès International v. Rothschild (2023), decided before the Southern District Court of New York and the Second Circuit Court of Appeals, determined that replication of the famous Birkin bag through NFTs constituted trademark infringement and dilution. The Courts brought these digital commodities into the conventional trademark discussion and determined that the same likelihood-of-confusion test applies to them in the virtual world, just as it does in the real world. 

 

Rosetta Stone Ltd. v. Google, Inc (2012), Court of Appeals, District of Virginia, determined that there were factual questions as to whether Google’s AdWords system posed a risk of consumer confusion and contributed to trademark infringement by allowing counterfeiters to buy and use the plaintiff’s trademarks as keywords in advertisements. While this case predates generative AI, it lays down crucial precedents related to the liability of intermediaries in digital marketplace environments. With search engines, recommendation algorithms, and other technological solutions becoming an integral part of online commerce transactions, the reasoning in Rosetta Stone is very helpful in evaluating the liability of technology vendors for contributory trademark infringement in cases involving generative AI technologies. 

 

In 1-800 Contacts, Inc. v. Lens.com, Inc.(2013), the Court of Appeals for the Federal Circuit determined that purchasing another firm’s trademark for use in search engine advertising was not sufficient in itself to establish trademark infringement under the Lanham Act. The Court stressed that the issue is whether the use causes consumer confusion as to the source, sponsorship, or affiliation of the goods or services. Although decided before the emergence of generative AI, the decision provides an important framework for assessing AI-driven advertising platforms and algorithmic recommendation systems, where the automated use of trademarks must still be evaluated through the traditional likelihood-of-confusion analysis. 

 

In Tiffany (NJ) Inc. v. eBay Inc.(2010), the Court of Appeals for the District Court for the Southern District of New York, it was held that the mere selling of counterfeits on the website cannot make a website contributorily liable for trademark infringement. The Second Circuit has established that liability will attach when the website operator has actual knowledge of specific listings or sellers and does not take reasonable steps to remedy the situation. Even though this case was heard before the emergence of generative AI technology, it lays down an important legal principle for the analysis of AI-enabled e-commerce websites. 

HOW ARE BRANDS FACING DILUTION CHALLENGES?

Blurring and Tarnishing at Machine Speed    

Brand dilution is a separate cause of action from infringement. Whereas proof of consumer confusion is required for infringement, the dilution law can afford protection to a famous mark regardless of consumer confusion under the Lanham Act’s dilution provisions, a complainant must prove that the mark was diluted through tarnishment—the dilution of the mark into association with poor or unappealing products—or blurring—the loss of distinctiveness of the mark.  Both processes have been vastly amplified by AI. 

The E-Commerce Dilution Epidemic 

The brand dilution arena at scale is now the online marketplace. In an e-commerce market featuring millions of third-party sellers, multijurisdictional transactions, and global expansion, counterfeiting, unauthorized use, and brand dilution are commonplace. The rapid expansion of virtual products and online trading has created novel methods for trademark abuse, making the potential risk of misuse, infringement, and trademark dilution greater than ever before. 

To counter this, key platforms have created AI-driven infrastructure for brand protection. Amazon is using machine learning algorithms through their Brands Registry feature to flag and block trademark-infringing listings before they adversely impact brand equity. The challenge is that counterfeiters and brand infringers are also using AI to stay one step ahead, creating new product images, modifying product listings that bypass image recognition software, and operating in countries or regions with weak legal enforcement to handle product deliveries. The result is an arms race in brand protection in which both parties use machine learning to gain an edge.

ENFORCEMENT CHALLENGES: LEGAL, TECHNICAL, AND JURISDICTIONAL

The Confusion Standard in a Post-Human Purchase Environment

The likelihood-of-confusion test is central to most trademark infringement cases and is evaluated from the perspective of the average consumer. This test is struggling to keep up with the changing business environment powered by AI. The traditional definition of the average consumer is one with less than flawless memory who cannot envision both products side by side. But increasingly, the purchase decisions /patterns of humans are being analysed by AI assistants in devices; these systems can quickly compare products and process large amounts of product information and reflect it on social media platforms. 

The doctrinal issues that have begun to be encountered by courts arise naturally: how does the likelihood of-confusion test apply to a “consumer” that is not human? Legal academics have commented that confusion is the pivot upon which trademark law turns, and a change to the confusion standard for AI purchasing agents may extend trademark liability throughout the digital marketplace in ways that were not contemplated by the Lanham Act’s drafters.

Attribution, Liability, and the AI Intermediary 

A conundrum associated with the practical enforcement of this situation is the question of liability in the case of an infringing AI system. The old-fashioned approach to trademark protection targets bad actors who are identifiable—human or corporate. The AI context complicates this: a user gives a prompt, a model produces content based on its training, the content is published on a platform or set of platforms, and a third-party purchases, reproduces, or publishes it; attributing infringement and thus liability is disputed at every stage. 

The issue of whether an AI developer should bear liability for infringing AI output remains open for debate. Various theories have been suggested by courts and scholars, and contributory trademark infringement comes into play when it is determined that the developer was knowingly involved in an act of trademark infringement.

Cross-Border Enforcement in a Borderless Digital Market 

While trademarks are primarily a territorial concept, the internet is not. The combination of an infringing listing generated by an AI located in one jurisdiction, hosted on a server in a second jurisdiction, and offered by a seller in a third creates a multijurisdictional enforcement issue that lacks a complete solution in any single legal jurisdiction. China has a major presence in both manufacturing and e-commerce activities across the world. From an American standpoint, a look at China may be imperative. In 2024, China made major proposals to its trademark law that would include a requirement of proof of use to maintain trademark registrations and explicit bad-faith provisions in existing law.  

CONCLUSION 

The essence of trademark law is to ensure that market information is coherent—that consumers can depend on brand names linking to the origin of products, and that brand owners can benefit from the reputation earned by developing quality products and services. Such is the nature of AI that poses significant challenges to both these functions: generating AI-drafted content that impersonates protected marks and operating at speeds and scales that severely outpace traditional enforcement endeavors. 

The courts have been actively engaged with these changes, applying established trademark concepts and doctrines to matters related to AI-generated content and digital or virtual goods. AI-driven systems are a major focus of investment for platforms. The structure of trademark law is being questioned at its foundational level by legislators and regulators as AI poses new challenges. However, the gap between legal adaptation and changing technology remains large. 

In the era of AI, a compelling imperative exists for brand owners: trademarks must be continuously monitored and reviewed, a strong digital enforcement program must be maintained, and early engagement with legal experts who possess not only IP expertise but also an understanding of the cutting-edge technology behind these systems is essential. For legal regimes, the problem does not lie in the rejection of the basic premises on which trademark law stands, but rather in the adaptation of those same principles to a marketplace increasingly influenced by AI technologies and digital commerce.

REFERENCES 
  1. McCarthy, J. Thomas. McCarthy on Trademarks and Unfair Competition (5th ed.). Thomson Reuters/West Publishing. Available at Thomson Reuters.  
  2. United States Congress. Lanham Act (Trademark Act of 1946), 15 U.S.C. §§ 1051 et seq. Legal Information Institute, Cornell Law School. Available at: Legal Information Institute – Lanham Act.  
  3. Cornell Law School, Legal Information Institute. Trademark – Wex Legal Dictionary/Encyclopedia. Published by the Legal Information Institute, Cornell Law School. Available at: Trademark (Wex).  
  4. World Intellectual Property Organization (WIPO). Artificial Intelligence and Intellectual Property. Published by WIPO. Available at: WIPO – Artificial Intelligence and Intellectual Property.  
  5. Hermès International v. Rothschild, 590 F. Supp. 3d 647 (S.D.N.Y. 2023), aff’d by the U.S. Court of Appeals for the Second Circuit. Official court records and reported judgments.  
  6. Beebe, Barton. An Empirical Study of the Multifactor Tests for Trademark Infringement. California Law Review, Vol. 94, No. 6 (2006). University of California, Berkeley School of Law.  
  7. Lemley, Mark A., and McKenna, Mark P. Irrelevant Confusion. Stanford Law Review, Vol. 62 (2010). Stanford Law School.  
  8. Dogan, Stacey L., and Lemley, Mark A. The Merchandising Right: Fragile Theory or Fait Accompli? Emory Law Journal, Vol. 54 (2005). Emory University School of Law.  
  9. Amazon. Amazon Brand Registry. Published by Amazon. Available at: Amazon Brand Registry 

Authored by: Manan Jhamb; Edited by: Tanmay Dhiman  

This article has been authored by an ultimate BBA-LLB student at the University Institute of Legal Studies, Chandigarh University. It was created during their legal internship tenure with us.