OpenAI’s brand protection is understandable. ThreatGPT, MedicalGPT, DateGPT, and DirtyGPT are just a few companies that have applied for USPTO trademarks recently.
All are riding the success of ChatGPT, OpenAI’s November chatbot based on its deep learning model, GPT-4, released last month.
After registering for a trademark for “GPT,” short for “Generative Pre-trained Transformer,” in late December, OpenAI this month petitioned the USPTO to speed up the process, claiming the “myriad infringements and counterfeit apps” emerging.
OpenAI lost its petition last week. The commission said OpenAI’s counsel failed to pay a fee and present “appropriate documentary evidence supporting the justification of special action.”
Jefferson Scher, a lawyer at Carr & Ferrell’s intellectual property division and chair of its trademark practice group, believes a decision may take up to five more months, given OpenAI’s current backlog. Scher says it’s still uncertain.
He argues OpenAI has good cause to expect the patent. We asked him if OpenAI would meet opposition because the “T” in GPT stands for “Transformer,” a neural network design initially introduced by Google researchers in 2017 and now widely used. Scher wonders if c may be a brand despite its descriptive roots. He cites IBM, short for International Business Machines, as an example of a brand with a descriptive beginning, even if poor. Scher notes that there is “no guarantee [OpenAI] could end up owning [GPT],” but precedents assist.
Scher also finds it beneficial that OpenAI has been utilizing “GPT” for years, having released its first Generative Pre-trained Transformer model, GPT-1, in October 2018.
Scher noted that it’s a “funny situation” because “usually when you’re basing claim on use, you have gradually built up your brand in the marketplace.” Still, OpenAI was mostly known to AI researchers until last year. Then, it released a mesmerizing deep learning model that generates digital images (DALL-E 2), followed by ChatGPT, making it an overnight sensation.
Even if a USPTO examiner approves OpenAI’s registration, it will enter an opposition phase when other market participants might argue against the “GPT” trademark.
Scher says: OpenAI would have to prove that “GPT” is proprietary and that the public understands it as such, rather than assuming it refers to generative AI.
How would the USPTO decide public perception? Scher suggests randomly sampling Americans and asking them to answer the question. However, that’s a six-figure effort the government won’t fund, so any OpenAI competitor would have to pay for it.
From late-night chat programs to public writing, “GPT” has been employed to shape public opinion. “If people aren’t treating it as proprietary, then a trademark trial would decide if it’s protectable,” adds Scher.
That would be a lengthy process, which OpenAI certainly doesn’t want.
The corporation should have protected “GPT” earlier. Instead, Scher believes the corporation was “probably caught off guard” by its success. In China, where it has not launched ChatGPT and may not be authorized, it sought to register a comparable trademark.
Scher believes “we’ve crossed a line where GPT is not three random letters.” I would tell a startup not to embrace it.
Scher argues trademark law’s renowned element may favor OpenAI. Famous brands are protected beyond their industry, yet anybody may get a trademark. For example, Rolex is too well-known to utilize other products. So even though it’s costly to pursue violators, OpenAI can restrict the widespread use of “GPT” if it can prove it’s a renowned trademark.
The firm may benefit from this lengthy procedure. OpenAI gains users and coverage as time passes, making that last possibility more realistic.
Does the typical household or person know about OpenAI? Scher adds, “They may be close.”
