OpenAI’s trademark application for “GPT” was refused by the USPTO because the word is “merely descriptive.” OpenAI’s identity is hurt, but don’t expect its rivals to release their own chatbots.
ChatGPT, the most popular conversational model on the market, is the most well-known AI brand and the one that pushed huge language models from curiosity to a worldwide trend.
However, the USPTO says the name does not match trademark registration requirements and does not qualify for “TM” protection. The application was denied once in October, and this is the “final” rejection.
According to the denial:
The mark indicates a feature, function, or attribute of the applicant’s products and services; hence, registration is denied.
OpenAI said it popularized the acronym GPT, which stands for “generative pre-trained transformer,” to describe the machine learning approach. It’s generative because it generates fresh (ish) stuff, pre-trained because it’s a huge model trained centrally on a private database, and transformer because Google engineers found a way to train considerably larger models in 2017.
According to the patent office, GPT was previously used in many other situations and by other firms in comparable ones. Amazon describes GPTs and their usage.
The patent claims GPT identifies a product feature. Say you had Crunchy O’s cereal and attempted to copyright “crunchy.” ChatGPT is a GPT-type AI model you speak with, which OpenAI did not design and is not alone in delivering. Though recognized, it doesn’t fulfill trademarking standards.
Without a trademark, OpenAI may lose its GPT-related word supremacy. Look for “TalkGPT,” no related, in app stores (they’re already there and many), and OpenAI can’t sue them for using their brand.
Although their legal protections are limited, OpenAI has the largest mindshare when someone mentions “GPT,” so they have the first-brand advantage. They may even increase GPT branding, trademarks aside, to prove OpenAI was first.
OpenAI has no ‘GPT’ trademark.
