Click the topic to read previous chapters:
Chapter VI Trademark Protection of AI
Artificial intelligence (AI) is currently a hot field in China. Most of the major domestic manufacturers such as Alibaba, Tencent, and Baidu, foreign giants such as IBM and MICROSOFT, or start-up technology companies favored by capital involve the AI field, and some of the latter have developed into emerging unicorn companies.
At present, most articles on the protection of AI in the field of intellectual property rights focus on patents, copyrights, and trade secrets, while the protection and influence of AI on trademarks are very rarely discussed. We try to analyze the early layout and later protection of AI-related trademark in this chapter, hoping to help enterprises.
6.1 Early Layout of AI Trademark
No matter in most countries and jurisdictions represented by China that determine the ownership of trademark rights mainly relying on “first filing”, or in other countries represented by the United States that determine the ownership of trademark rights mainly relying on “first use”, it is necessary to file national trademark applications or designate the mark through the Madrid System for the International Registration of Marks as early as possible, so as to occupy an earlier filing date.
Before filing an AI trademark application, there are two aspects that need to be determined. One is to select an appropriate AI trademark, and the other is to specify the corresponding goods and services.
The problem with selecting an appropriate AI trademark is that it is becoming more and more difficult to select an appropriate AI trademark without prior conflicts, especially when most applicants hope to select the same AI trademark in all markets around the world, considering on the one hand, applicants preferring to select concise and catchy trademarks with harmonious meaning and hi-tech indication, and on the other hand, the existing stock and annual huge increase of trademarks in various countries and regions.
We have noticed that under such circumstances, major domestic manufacturers and foreign giants will choose to apply for new AI trademarks by combining existing main trademarks with new catchy names/nouns, such as IBM’s IBM Watson, MICROSOFT’s MICROSOFT AI SOLUTION, Huawei’s HUAWEI ATLA, Alibaba’s Alibaba Cloud, and Tencent’s Tencent Cloud. The advantage of such choice is that on the one side, the previous main trademark registered for many years can greatly ensure the chances for the new combined trademark to pass the examination smoothly, and are not likely to collide with the other existing trademarks, and on the other side, consumers will have trust on the goods/services bearing such combined marks knowing from which manufacturers the goods and services are from, and meanwhile know the trademarks are for new series of production line due to the new name/noun behind the main trademark, and the newly added nouns like “Cloud” and “AI SOLUTION” can even function to show the specific contents of the goods and services.
However, due to various considerations, some major manufacturers choose to adopt brand-new AI trademarks that are completely unrelated to the previous main trademarks, such as Baidu’s Apollo. The advantage of such choice is that the brand-new brand will give consumers a refreshing impression and a more concise appellation; plus, such brand-new brands grant more flexibility of trademark use in actual commerce, that is, after registration, it can be used alone as well as in combination with the main trademark based on the circumstances. The problem is, it greatly increases the risk of colliding with the prior trademark, even for many large companies with fairly complete IP strategies. For example, when Huawei tried to apply for the registration of its AI trademark “HiAI”, it was found that its partner had already filed the same application dozens of days ago, and then two parties fought for this trademark for several years; another example is when AMAZON tried to apply for the registration of its AI trademark “ECHO” in China, it also encountered prior obstacles and failed to have it registered for the moment. Therefore, if choosing a brand-new AI trademark, the company must firstly do a good job in confidentiality work and sign a confidentiality agreement with the partners, secondly, do a lot of retrieval beforehand to evaluate the possibility of registration, and thirdly, file an application as soon as possible to get an early filing date. But even if it takes a huge amount of time and economic cost, sometimes the registration result of such a brand-new AI trademark is not satisfactory, especially if the company intends to register and use it globally.
Emerging AI technology companies do not have such long-standing main trademarks to rely on, and are often at the stage of choosing their own main trademarks for AI products and services. In order to avoid too many prior conflicting trademarks when choosing a brand-new main trademark, try to choose a coined word rather than dictionary vocabulary. If the coined word can have elements corresponding to its company name and can imply the attributes of a technology company, the trademark will have a better performance in terms of registration pass rate, marketing convenience and potential consumer acceptance. IFLYTEK, the English trademark of iFLYTEK, is one good example. Under the influence of trademarks such as IPHONE, IPAD, and IWATCH, consumers can easily understand ‘I” as an element representing new high-tech technologies; the “TEK” is homophonic to “TECH”, which also implies the element of technology; for one thing, the dictionary vocabulary “FLY” inserted in the middle corresponds to the meaning of “飞” in the company name, and for another, the addition of the preceding and following “i” and “TEK” elements will not cause conflicts with other trademarks containing the separate word “FLY”.
In addition to the careful selection of AI trademarks, the goods and services designated by AI trademarks also need to be carefully considered. Different from other intellectual property rights, the registration and protection of trademarks must be combined with the corresponding goods and services. Before the AI trademark application is filed for registration, the designated goods and services must be selected. At present, most countries and jurisdictions in the world adopt the Nice Classification, which divides common goods and services into 45 major classes. The essence of AI technology is a software program that can implement various functions, so its core protection category is Class 9 in the Nice Classification; at the same time, the upgrade and maintenance of the software program belongs to Class 42 in the Nice Classification.
However, different from other technologies, AI technology can also be widely used in other fields in addition to its products and services. From vehicle & driving to lighting & cooking, from living & home furnishing to toy & education, you always can see the application of AI technology. Although in many cases, the AI technology owner just provides AI technology for another manufacturer for use in its vehicles, lamps, kitchenware, household supplies, toys, and online education services, instead of producing and selling such products or providing such services by itself; but if the owner of AI technology fails to apply for the registration and protection of its own trademarks in the corresponding Classes 12 (vehicles), 11 (lamps and kitchenware), 21 (household supplies), 28 (toys) and 41 (education services), other parties may register highly similar or even identical trademarks with AI trademarks on these goods and services, making consumers easily think that the technology of the AI technology owner is used in these goods and services, thereby resulting in confusion and misleading, and even a bad impact on the goodwill of the AI technology owner. So should the AI trademark cover the goods and services in all of these possible application areas when selecting goods and services in registration application? The answer is not absolutely so.
As mentioned above, in most countries and jurisdictions represented by China, the ownership of trademark rights is determined mainly relying on “first filing”, while in some countries represented by the United States, the ownership of trademark rights is determined mainly relying on “first use”. Under the former system, trademark applicants can apply for registration of and maintain a trademark without necessity to voluntarily provide evidence on trademark use; while under the latter system, trademark applicants can apply for registration of and maintain a trademark only after voluntarily providing evidence on trademark use. Due to the difference of trademark systems, AI technology owners can consider covering all the above related categories when designating goods and services in addition to the core categories in the former countries and jurisdictions so as to prevent others from registering the same and similar trademarks on vehicles, lamps, kitchenware, household supplies, toys and online education services, causing consumers to believe that such goods and services have the AI technology of the AI technology owner; while under the latter system, especially in the United States where false statements and evidence on trademark use will result in trademark invalidation, fine and imprisonment, it is recommended that AI technology owners apply for registration only in the core categories that they actually use.
Comparing the applications for AI trademark registrations by the above-mentioned companies such as Alibaba, Tencent, Baidu, IBM, MICROSOFT, and iFLYTEK under the two systems, we have indeed found that they often cover all relevant categories in the Nice Classification as more as possible under the former system, for example, in China; and under the latter system, for example, in the United States, often only designate the actual used categories in the Nice Classification, less than those designated in China.
Therefore, the selection of designated goods and services for AI trademark requires comprehensive consideration of local trademark practices, and it is important to avoid “one-size-fits-all approach” in all countries and jurisdictions, resulting in insufficient protection under the former system, or designating goods and services that are not actually used under the latter system, which leads to the subsequent failure of the registration and maintenance of and invalidation of trademark, or fine and other severe penalties.
6.2 Later Protection of AI Trademark
The most fundamental purpose of trademark registration is to protect one’s brand; on the one hand, to make consumers understand the AI trademark and which manufacturer its goods and services come from, and on the other hand, to prevent other parties from registering and using the same and similar trademarks on the same and similar goods; the purpose of the AI trademark registration is nothing more than this.
Similarity can be easily judged when the trademarks and goods are identical. However, when the trademarks and goods are different, the practices of different countries and jurisdictions are different in judging whether the trademarks and goods are similar. Such a situation inevitably produces confusion as much as trademark confusion. Generally speaking, China’s practice tends to make judgment in strict accordance with the Similar Goods and Services Table, while the United States will take into more consideration of the influencing factors in actual business. Moreover, even in the same country and jurisdiction, the judgment standards and consideration factors are not the same at the stage of administrative right grant and in civil tort cases.
For example, at the early stage of administrative right grant on trademark, the United States use The Polaroid Factors determined in the Polaroid Corp. vs. Polaroid Elecs, Corp. to determine whether the trademarks of both parties constitute similar trademarks on similar goods and services. The Polaroid Factors include:
1. the strength of the senior user’s mark;
2. the degree of similarity between the two marks;
3. the proximity of the products;
4. the likelihood that the prior owner will bridge the gap;
5. actual confusion;
6. the junior user’s good faith in adopting its own mark;
7. the quality of junior user’s product; and
8. the sophistication of buyers.
In the later judgment of civil infringement, the US court followed the John H. Harland Co. vs. Clarke Checks, Inc., in which the following seven elements need to be considered in determining whether the trademark infringement is constituted, namely:
A. Type of Trademark.
B. Similarity of Design.
C. Similarity of Products.
D. Identity of Retail Outlets and Purchasers.
E. Similarity of Advertising Media Used.
F. Defendant's Intent.
G. Actual Confusion.
It can be seen from the comparison that the two judgment standards are mostly overlapped, but the former considers the strength of trademarks and public interests more, while the latter is more inclined to the scenario of actual sale.
However, what is gratifying is that in current Chinese judicial practice, considering the specific application scenarios of AI, various influencing factors in actual commercial use have been taken into consideration more rather than determining the similarity of trademarks and goods, and then drawing a conclusion on whether the trademark infringement is constituted by strictly following the Similar Goods and Services Table. In the Baidu Online vs. Beijing Zile Technology Co., Ltd. (2021), the defendant prominently referred to its products as “Xiaodu” in the Duyaya learning machine produced and sold; however, products such as smart speakers and learning machines using AI technology in the market, consumers call products more often with voice rather than traditionally relying more on text to identify products, which will lead to the fact that “Xiaodu (小度)” and “Xiaodu (小杜)”, which do not generally constitute similar trademarks based on Chinese text, are easily confused and misidentified in the application scenario of such AI technology. Considering the popularity of the plaintiff’s “Xiaodu (小度)” and “xiaodu xiaodu(小度小度)” and the defendant’s subjective bad faith, the People’s Court of Haidian District, Beijing judged beyond the Similar Goods and Services Table that the plaintiff’s Xiaodu smart speaker and the defendant’s Duyaya learning machine are similar products in terms of functions, audiences, and sales channels.
In addition to voice call interaction, when consumers use keywords to find suitable products and services during shopping on e-commerce platforms, and search for information about corresponding products and services in search engines, AI technology is also increasingly being used. With the sales of e-commerce platforms beginning to surpass offline stores, and search engines providing consumers with more product and service information than traditional media, AI technology is playing an increasingly critical role in consumers’ purchasing decisions. In this case, whether AI technology providers will become potential trademark infringers is a question that every AI technology provider needs to think about.
In the Coty Germany GmbH vs. Amazon Services Europe Sàrl, Amazon Europe Core Sàrl, Amazon FC Graben GmbH, Amazon EU Sàrl (2020), Amazon’s AI technology did not automatically select keywords and was not actively involved in the keyword advertising system, so it is determined that Amazon shall be exempted from liability for trademark infringement. Therefore, if AI technology does not automatically and selectively “help” consumers choose keywords, nor is it actively involved in the keyword advertising system of the seller in question, and the AI technology provider will also implement corresponding removal procedures later and timely follow up the implementation, the AI technology provider shall be exempted from liability for trademark infringement.
However, if the AI technology provider is involved in more infringements, it may constitute trademark infringement. In the Cosmetic Warriors Ltd an Lush Ltd vs. Amazon.co.uk Ltd and Amazon EU Sarl (2014), the defendant Amazon was ruled to bear corresponding legal liabilities since the website links triggered by Amazon’s AI technology after consumers use the trademark search do not contain the brand products corresponding to the trademark, and consumers may be confused and even misled as whether the products sold on the website are from the brand owner of the targeted trademark.
In summary, AI technology providers need to select appropriate trademarks according to their own circumstances in the early trademark layout, and consider different trademark practices in different countries and jurisdictions to designate different goods and services; in the later trademark protection, on the one hand, they need to decide whether to take rights protection actions with a comprehensive consideration of the special features of AI technology application scenarios while protecting their own rights and interests, and on the other hand, they need to pay attention to whether their own algorithm technology has been excessively and inappropriately involved in consumers’ purchasing decisions.