If startup OpenAI has felt protective about its brand lately, that’s understandable. ThreatGPT, MedicalGPT, DateGPT, and DirtyGPT are just a sample of the many organizations that have applied for trademarks with the USPTO in recent months.
They all fall back on the incredible popularity of ChatGPT, the chatbot introduced in November by OpenAI that’s built on the company’s deep learning model, and the latest version of which, GPT-4, was released last month.
It’s no wonder that after filing in late December for a trademark for “GPT,” which stands for “Generative Pre-training Transformer,” OpenAI last month petitioned the USPTO to speed up the process, citing “countless violations and fake applications” that Spring begin to exist.
Unfortunately for OpenAI, its petition was denied last week. According to the agency, OpenAI’s attorneys neglected to pay the relevant fees as well as provide “appropriate documentary evidence supporting the justification for the private action.”
Given the rest of the waiting list that OpenAI finds itself in, that means a decision could take up to another five months, says Jefferson Scheer, partner in Carr & Ferrell’s intellectual property group and head of the company’s brand practice group. Even then, Scheer explains, the outcome is uncertain.
Certainly, he says, OpenAI has plenty of reason to expect it will be able to secure the patent. We asked him, for example, if OpenAI might encounter resistance given that the “T” in GPT stands for “Transformer,” the name of a neural network architecture first unveiled by Google researchers in 2017 that has come into widespread use. . “Can GPT be a trademark even if it has a very descriptive origin?” Cher asks. He could point, he says, to IBM, short for International Business Machines, as just one example of a trademark having a descriptive etymology, even if the description is poor. This “doesn’t guarantee that (OpenAI) may end up owning (GPT),” Sher adds, but such precedents help.
Also helpful, says Scherr, is the fact that OpenAI has been using “GPT” for years, having released its original generative pre-trained transformer model, or GPT-1, in October 2018.
Again, though, Scheer noted that it’s a “funny situation,” where “typically, when you base on a usage claim, you build your brand incrementally in the market,” whereas OpenAI was known primarily to AI researchers until Last year, when the release of an amazing deep learning model that generates digital images (DALL-E 2), followed by ChatGPT, turned the company into a kind of overnight sensation.
Even if a USPTO examiner does not have an issue with an OpenAI application, it is then moved into what is called an opposition period, during which other market participants can discuss why the agency rejected the “GPT” trademark.
Scherr describes it this way: in the case of OpenAI, it would have to prove that “GPT” is a property and that the public perceives it as such rather than perceiving the acronym related to generative AI more broadly.
How will the USPTO pass its verdict on public perception? “One scenario is you take a random sample of Americans and ask them to answer the question,” Scherr says, but this is a six-figure project the government won’t pay for, so any competitor to OpenAI needs to bill something like that.
Another means is to establish the connection of public perception of how “GPT” is used in public, from late-night talk shows to public writing. “If people don’t treat it as property, a trademark trial will decide whether or not it’s protectable,” Scherr says.
Not surprisingly, this entailed a lengthy process, which is definitely the last thing OpenAI wants.
It raises the question of why the company didn’t move to protect GPT sooner. Here, Scherr speculates that the company “may have been surprised” by its success. (In fact, it appears to be trying to get ahead of things in China, where ChatGPT hasn’t launched yet and may not be allowed to, but it has tried to register a related trademark.)
Either way, Scherr says, he finds that “we’ve crossed a line where GPT isn’t just three random characters. If you (a startup) asked me if it was safe to adopt it, I’d say it’s not safe.”
OpenAI may benefit most from an aspect of trademark law where the goodwill of a registered trademark is a dominant factor, Scherr says. While one does not need to be famous to secure a trademark, once an outfit becomes famous, it receives protection, even outside its field. Rolex is too famous a brand to be used for anything else, for example. If OpenAI can prove that “GPT” is a well-known trademark, the company will also be able to prevent widespread use of the acronym (even if it would be costly to prosecute offenders).
That could be the positive side for the company in this long process. The more time passes and the number of users of OpenAI increases and the more coverage the company gets, the more likely the latter scenario will occur.
Is OpenAI known in the average household, by the average person? “Sure, they might come close,” Scherr says.