[Guest post] Benelux Office rules on opposition proceedings prepared by ChatGPT

Artificial intelligence (‘AI’) has surely been the hot topic of the last two years. In similar vein to a quote from a 1999 film, it seems that “AI is everywhere and all around us”.

One of the more interesting aspects of AI is the use of AI generated output, by tools such as ChatGPT or OpenAI. Stories of legal professionals relying on case-law cited by ChatGPT, only to later find out such case-law in fact does not exist, are known all too well.

But what about legal proceedings where there is no human involvement at all? As unthinkable it may seemed just years ago, the rapid developments in the field of AI could result in a ‘trial by AI’ sooner than we think.

A recent decision of the Benelux Office of Intellectual Property should be of particular relevance to those interested in AI, giving a unique insight how AI generated legal arguments hold up in trade mark opposition proceedings.

The case at hand concerned opposition proceedings initiated by Penguin Books Limited as the opponent. Penguin Books is the proprietor of the renowned trade marks for books and literature. In the opposition proceedings before the Benelux Office, it relied on several EU trade marks (‘EUTMs’), two of which contained the word element ‘PENGUIN’. The marks were all registered for Class 16 (books) of the Nice Classification.

The defendant in this case is a private person and the applicant of the Benelux trade mark ‘ARTPENGUIN’, which was also applied for goods of Class 16. The mark is depicted below:

Having more confidence in AI representation than human representation, the defendant indicated that it was represented by ‘Chat GPT (OpenAI’) and that it fully relied on arguments generated by AI output.

With respect to the opposed mark, Penguin Books claimed that, due to ‘PENGUIN’ being part of the word element ‘ARTPENGUIN’, resulting in aural, conceptual and visual similarities, there would be a risk of confusion within the meaning of article 5 Trade Mark Directive.

The applicant then requested Chat GPT to prepare arguments in its defence, which were subsequently filed with BOIP. The AI output produced three legal arguments. Firstly, ChatGPT rejected there could be visual similarity, since the penguin device of the opposed mark was ‘modern’, in contrast to the rather “old school” device of the Penguin Books trade marks. Secondly, it held that there could be no aural similarity, due to the element ‘ART’ in the opposed mark. Lastly, Chat GPT alleged that the visual difference (modern v. old school) would also render the marks to be conceptually dissimilar, which would lead to a difference audience being drawn to it.

The Benelux Office was only mildly impressed by Chat GPT’s legal output. The Office first established that ChatGPT did not address the matter of similarity of the goods at issue, and in absence of raising a defence in this regard, the Benelux Office ruled that the goods were indeed similar.

Moving on to conceptual similarity, the Benelux Office found that, due to the opposed mark containing the word element ‘penguin’ and depicting an actual penguin, there was an ‘undeniable’ conceptual similarity, since both marks allude to a penguin.

In assessing aural similarity, the Benelux Office, not without a sense of humour, noted that ChatGPT “could have been given a better prompt” by the defendant, since the latter had included a mark that was not part of the opposition (the word mark ‘Penguin Books’). In assessment of the marks that are part of the opposition, the Benelux Office found that, despite the element ‘ART’, the marks both contained the element ‘PENGUIN’ and were therefore similar to a strong degree.

AI-generated “Kat using ChatGPT”

Finishing with the visual similarity, the Benelux Office recalled that in a figurative mark with word elements, the word elements will usually attract more attention from the relevant public. Therefore, despite accepting that there are some visual differences in both marks, the Office concluded that the marks were at least visually similar to an average degree.

As its final ruling, taking into account the identical goods and the conceptual, aural and visual similarity, the Benelux Office allowed the opposition and rejected the opposed mark for the goods of Class 16.

Is this the beginning of the end for legal representation and a new era of AI representation and trial by AI? This Katfriend is not (yet) convinced. Although AI has the potential to be a great help when drafting legal arguments, it should not be the sole legal source to rely on. As evidenced in the decision, mistakes are easily made and the AI generated output can be inaccurate, irrelevant or of a subpar quality.

Still, it is interesting to see that an AI program was able to come up with arguments that were neither frivolous nor could be dismissed without legal examination. As such, the future appears to be closer than we might think (or want to accept) …