What Happens If You Use ChatGPT-Generated Images to Create LINE Stickers or Apply for a Trademark?

In recent years, AI technology has advanced rapidly. Generative AI tools like ChatGPT can not only create written content but also generate high-quality images, benefiting many businesses and creators. However, when planning to use AI-generated images for commercial purposes—such as creating and selling LINE stickers[1] or applying for trademarks—what legal and practical issues might arise?

Check for Commercial Use Restrictions in Generative AI Terms of Service

Before using AI-generated images for commercial purposes, such as making LINE stickers, applicants should first verify whether the terms of service for the AI tool permit commercial use, to avoid violating licensing agreements.

For example, companies like OpenAI and Adobe allow commercial use of images generated by tools like ChatGPT and Firefly, but remind users to avoid infringing on third-party rights. Midjourney distinguishes between paid and free users: only paid members (especially corporate users) obtain ownership of the generated images. However, it does not clearly state whether commercial use is permitted[2].

Additionally, AI-generated images are generally not protected under copyright law because they are not considered human creations. Nevertheless, if the image meets trademark law requirements, users may still apply for trademark registration and enjoy exclusive rights.

Basic Trademark Concepts and the Requirement of Distinctiveness

The primary function of a trademark is to identify the source of goods or services and prevent consumer confusion. When applying for a trademark, the applicant must ensure that the mark is distinctive to qualify for registration. Distinctiveness can be inherent or acquired through long-term use. Inherent distinctiveness typically falls into the following categories[3]:

Fanciful Marks (Very High Distinctiveness): These are invented terms that do not exist in common language and serve solely to distinguish a product or service (e.g., "Google" for search engine services).
Arbitrary Marks (High Distinctiveness): These use existing words unrelated to the product or service they represent (e.g., "Apple" for electronic products).
Suggestive Marks (Moderate Distinctiveness): These imply qualities or characteristics of a product or service indirectly and require imagination to interpret the connection (e.g., "Netflix" suggests internet + films).
Descriptive Marks (Low Distinctiveness): These directly describe a feature or characteristic of the product or service (e.g., “Express,” “BBQ,” or “Luxury”). Such marks are generally difficult to register.
Generic Marks (Very Low Distinctiveness): These are common terms or names for goods or services (e.g., “Computer” for computers or “Arabica” for a coffee variety), which cannot be registered.

If an applicant wishes to register an AI-generated image as a trademark, the image must have sufficient distinctiveness, not resemble existing trademarks, and must not mislead or confuse consumers.

The ChatGPT Trademark Application Was Rejected by the USPTO

It is important to note that trademark protection is based on territorial principles—trademark rights are limited to specific countries or regions. For example, OpenAI attempted to register “ChatGPT” as a trademark in the United States, but the U.S. Patent and Trademark Office (USPTO) rejected the application. The USPTO held that "CHAT" means synchronous communication over a computer network, and "GPT" is a widely used acronym for “Generative Pre-trained Transformers,” a type of neural network model that generates human-like content. As such, "ChatGPT" was deemed a descriptive mark and not eligible for trademark registration. The application was officially rejected on February 9, 2024[4].

However, in Taiwan, “ChatGPT” was successfully registered as a trademark[5], illustrating how different jurisdictions may have varying standards for examining trademarks and consumer perception of marks. Thus, it is essential to understand the local legal framework and market conditions when applying for trademarks in different regions.

Risk of Trademark Similarity in AI-Generated Images

A key requirement for trademark registration is distinctiveness, which allows consumers to distinguish one brand from another. If an AI-generated image is too generic or similar to an existing trademark, it may lack distinctiveness and be rejected. Additionally, since generative AI tools are trained on large datasets, there is a higher chance that users may unintentionally create marks similar to existing trademarks, increasing the risk of trademark similarity.

Risk of Trademark Disputes Involving AI-Generated Images

If others use the same AI model to generate similar images, trademark ownership conflicts may arise, potentially confusing consumers. This could make it difficult for the applicant to maintain exclusive rights to the image, weakening legal protection. Even after successful registration, AI-generated trademarks may be challenged by third parties.

Under Taiwan’s Trademark Act, if a third party finds a registered trademark similar to their own or believes it resembles materials in the public domain, they may take the following legal actions:

Filing an Opposition (Article 48): Within 3 months of a trademark’s publication, any person (not limited to stakeholders) may oppose the registration based on grounds listed in the Trademark Act.
Trademark Invalidation (Article 57): Interested parties or examiners may apply for a review of the registration if it violates the Trademark Act.
Trademark Cancellation (Article 63): A registered trademark may be canceled if it is unlawfully used, unused for a period, becomes generic, lacks proper labeling, or misleads the public. Any person may file for cancellation.
Trademark Infringement Lawsuit (Articles 68, 95–97): If the registered trademark is used in the market and causes harm to a third party, the injured party may seek civil damages or file a criminal complaint to protect their rights.

Conclusion: Challenges and Recommendations for AI-Generated Trademark Applications

Currently, using AI-generated images in trademark applications exists in a legal gray area. Applicants should carefully assess the distinctiveness of the image and understand usage limitations, while considering the risks of similarity and infringement.

To minimize legal risks, businesses and creators are advised to:

Manually modify and adjust the AI-generated images to ensure uniqueness and a distinctive style.
Use reverse image search tools to check for similarity with existing trademarks. Taiwan’s Intellectual Property Office now offers image-based trademark search capabilities[6].
Consult a trademark attorney to ensure the application complies with legal requirements.

As AI technology continues to evolve, legal systems will gradually adapt to such emerging issues. In the meantime, companies should be cautious when using tools like ChatGPT to generate images for trademarks.

References:

[1] ChatGPT Creates LINE Stickers! 3 Steps to Get It Done, Even the Free Version Works – Storm Media, April 2, 2025:https://www.storm.mg/lifestyle/5350186...

[2] Midjourney Terms of Service:https://docs.midjourney.com/hc/en-us/articles/32083055291277-Terms-of-Service...

[3] Taiwan Intellectual Property Office – Trademark Distinctiveness Examination Guidelines.

[4] USPTO, Office Action – Trademark Application Serial No. 97733261 – CHATGPT – OPAI015US.

[5] Registered Trademark No. 02341948 (Taiwan).

[6] TIPO Image-Based Trademark Search System: https://cloud.tipo.gov.tw/S282/S282WV1/