AI and Industrial Property: Trademarks Generated by Algorithms

This page is a translation of the original Portuguese text, provided for convenience. It deals with Brazilian law. If there is any discrepancy, the Portuguese version prevails. Read the original in Portuguese

Algorithms can already create logos, slogans, and complete visual identities, but Brazilian industrial property law has yet to define clearly who owns a trademark generated by artificial intelligence.

The Current Landscape of AI-Generated Trademarks

In recent years, we have seen a significant shift in how trademarks are conceived and developed. Generative artificial intelligence tools can now produce visual elements, business names, and even complete corporate identities in a matter of minutes, work that once took weeks of human creative effort. This technological revolution has raised a set of legal questions that still lack definitive answers under Brazilian law and in much of the world’s legislation.

When you look at how these tools work, you find that the algorithms are trained on vast databases containing millions of existing trademarks, logos, and graphic elements. From that training, the system learns aesthetic patterns, color combinations, and design trends. Its output can be surprisingly original or, in some cases, dangerously similar to trademarks that are already registered. That duality is precisely what makes the legal debate so complex and so urgent.

In Brazil, the Industrial Property Law (Law No. 9.279/1996) sets out the requirements for registering a trademark with the INPI (Brazil’s National Institute of Industrial Property). The sign must be visually perceptible and distinctive, and it must not fall under any of the statutory prohibitions. However, the law was drafted at a time when creating trademarks was an exclusively human activity, which leaves interpretive gaps when the creative process involves automated systems.

Ownership and Authorship: Who Owns a Trademark Created by AI?

The central question with algorithm-generated trademarks is ownership. Unlike copyright law (which protects artistic and literary works), trademark law does not require creative originality in the same sense. What it requires is distinctiveness. This means that, in theory, a trademark generated by AI could be registered as long as it meets the legal requirements. The problem is determining who qualifies as the legitimate applicant.

We see three main possibilities. The first assigns ownership to the developer of the AI tool, on the argument that the developer built the system capable of generating the mark. The second recognizes the user who operated the tool, supplying the parameters and instructions (the so-called “prompts”) that directed the creation. The third is more radical and still a minority view in academic debate. It would recognize some form of authorship in the system itself, something current Brazilian law does not allow, since the legal system requires the holder of rights to be an individual or a legal entity.

In practice, most generative AI platforms partially settle this question through their terms of use. Many of them transfer to the end user the rights to the results generated, provided certain conditions are met. Still, these contractual terms do not necessarily bind third parties, and they do not replace the review the INPI must carry out when it examines a registration application. In other words, a contractual layer sits on top of the regulatory layer, and the two do not always lead to the same conclusion.

Industrial property law was designed to protect human creations, and now we need to adapt it to a reality in which algorithms take an active part in the creative process.

Another relevant situation arises when multiple users, working with the same tool and similar parameters, end up with very similar visual results. In that scenario, distinctiveness becomes even more complicated, because two or more applicants could seek registration of substantially identical trademarks, both generated by the same algorithm. A trademark registration system built on the principle of priority would have to handle priority disputes over simultaneous, automated creations.

Risks of Conflict with Existing Trademarks

One of the most concrete problems we see with AI-generated trademarks is the risk of infringing third-party rights. Because the algorithms are trained on databases that include registered trademarks, there is a real possibility that the system will produce visual or word elements that infringe earlier registrations. The similarity may be unintentional (the algorithm does not deliberately “copy”), but the practical result can amount to reproduction or imitation of someone else’s trademark, which Brazilian law prohibits.

The risk shows up in different degrees. In the most obvious cases, the algorithm may generate a logo that nearly reproduces a famous trademark, and the unlawfulness would be easy to spot. In subtler (and more frequent) cases, the AI may produce combinations of colors, typefaces, and shapes that, although not identical to existing trademarks, are similar enough to confuse consumers. This gray area is especially problematic because the analysis of trademark conflicts relies on subjective criteria that vary with the market segment and the target audience.

On civil liability, the question is who answers for an infringement. Suppose a company hires an AI service to create its trademark and later learns that the result infringes a third party’s rights. Can it claim ignorance or good faith? We understand that the case law tends to hold liable whoever actually uses the trademark in the marketplace, regardless of how it was created. In other words, the fact that an algorithm generated the mark does not relieve its owner of the duty to run proper prior-rights searches before adopting it commercially.

We therefore recommend that companies and entrepreneurs who use AI tools to create trademarks conduct thorough prior-rights searches with the INPI before investing in the generated visual identity. Automating the creative process does not replace the legal diligence needed to make sure the chosen mark does not collide with existing rights. Skipping that step can lead to injunction suits to stop use of the mark, damages claims, and the loss of everything invested in building the brand.

Regulatory Outlook in Brazil and Around the World

Internationally, different jurisdictions have taken different approaches to the intersection of artificial intelligence and industrial property. The U.S. Patent and Trademark Office (USPTO) has published guidance reinforcing that only natural persons can be named as inventors on patent applications, and that the use of AI in the creative process must be properly disclosed. In the European Union, the AI Act sets transparency obligations for generative AI systems. This indirectly affects trademark creation by requiring that AI-generated content be identified as such.

In Brazil, the AI Legal Framework is still being worked out in the legislature, and its specific provisions on intellectual property remain under debate. We are closely following the discussions in the National Congress, which have addressed topics such as algorithmic transparency, liability for harm caused by AI systems, and rights over automatically generated content. There is still no consensus on whether industrial property legislation will need specific amendments or whether the existing rules can be interpreted broadly enough to accommodate the new technological realities.

The INPI, for its part, has shown growing interest in the subject. The federal agency has stated, at events and in institutional publications, that it needs to study the impact of AI on the industrial property system. Questions such as whether to use AI in examining registration applications (to identify conflicts more efficiently) and what criteria to apply to applications involving AI-assisted creations are on the horizon. We think it likely that in the coming years the INPI will publish specific guidelines on how to treat trademarks whose creation involved artificial intelligence systems.

The use of AI to monitor and defend trademarks also deserves attention. Large companies already use algorithms to scan the internet for improper uses of their marks. This practice raises questions about the limits of automated surveillance and about the risk that false positives will lead to abusive extrajudicial notices. We see AI acting as both a creation tool and an enforcement tool in the trademark ecosystem, expanding both the opportunities and the legal challenges.

Best Practices for Companies Using AI to Create Trademarks

Given where things stand, we have put together a set of practical recommendations for companies and professionals who want to use AI tools to create trademarks without exposing themselves to unnecessary legal risk. The first and most basic piece of advice is to treat the AI’s output as a starting point, not a finished product. Human involvement in selecting, refining, and validating the mark is essential, both to add distinctiveness and to show that a person contributed creatively to the process.

We also recommend that companies carefully document the entire creation process. That means recording the parameters used in the AI tool, the intermediate versions, the changes made by human designers, and the reasons for the final choice. This documentation can be decisive in any dispute over ownership or originality of the trademark. It is also prudent to review the terms of use of the AI platform, checking whether the rights to the generated results are actually transferred to the user and whether there are restrictions on commercial use.

A prior-rights search remains an indispensable step. Before filing a registration application with the INPI, we search both the agency’s database and international registries (such as the Madrid system, administered by WIPO, the World Intellectual Property Organization) to verify that the generated mark does not conflict with existing rights. This diligence protects the company against opposition proceedings, invalidity actions, and disputes that can jeopardize significant investments in branding.

Finally, we suggest that companies keep up with legislative and regulatory developments on the subject. The legal landscape for AI and intellectual property is changing quickly, and administrative decisions by the INPI or court precedents can significantly change the prevailing understanding. Having specialized legal counsel is essential to navigate this environment of regulatory uncertainty and to properly protect the company’s intangible assets.

This subject is directly related to intellectual property of, a topic we cover in a dedicated article.

To round out your reading, see also our article on decision algorithms.

As informações deste artigo são de caráter informativo e não substituem consulta jurídica individualizada. Cada caso possui particularidades que exigem análise específica.