‘GPT’ could soon be a trademark if OpenAI has its way

If startup OpenAI has been feeling protective of its brand lately, that’s understandable. ThreatGPT, MedicalGPT, DateGPT, and DirtyGPT are just a sample of the many outfits to file with the United States Patent and Trademark Office in recent months.

All building on the stunning popularity of ChatGPT, the chatbot deployed in November by OpenAI which is itself built on the company’s deep learning model, the latest version of which, GPT-4, was deployed on last month.

It’s no wonder that after applying in late December for a trademark for “GPT,” which stands for “Generative Pre-trained Transformer,” OpenAI last month petitioned the USPTO to expedite the process, citing the “myriad infringements and counterfeit applications” beginning to spring into existence.

Unfortunately for OpenAI, its petition was denied last week. According to the agency, OpenAI’s attorneys failed to pay associated fees and provide “appropriate documentary evidence to support the justification for special action.”

Given the rest of the queue OpenAI is in, that means a decision could take up to five more months, says Jefferson Scher, partner in Carr & Ferrell’s intellectual property group and chairman of the group. practice of the company’s brands. Even then, the outcome is not certain, says Scher.

Certainly, he says, OpenAI has many reasons to expect to be able to obtain the patent. We asked him, for example, if OpenAI might face resistance given that the “T” in GPT stands for “Transform,” which is the name of a neural network architecture that Google researchers unveiled for the first time in 2017 and has become widely used. . “Can GPT be a trademark even if it has a very descriptive origin?” Scher asks. He can, he says, point to IBM, short for International Business Machines, as a single example of a mark with a descriptive origin, even if the description is weak. It’s “no guarantee (OpenAI) could end up owning (GPT),” Scher adds, but such precedents help.

The fact that OpenAI has been using “GPT” for years also helps, Scher says, after releasing its original Generative Pre-Trained Transformer, or GPT-1, model in October 2018.

Again, however, Scher noted that this was a “funny situation”, in that “usually when you base a claim on use, you’ve gradually built your brand in the market” , while OpenAI was mostly known to AI researchers until the end. year when the release of a fascinating deep learning model that generates digital images (DALL-E 2), followed by ChatGPT, turned the company into something of an overnight sensation.

Even if a USPTO examiner has no problem with OpenAI’s application, it will then be moved to a so-called objection period, where other market participants can explain why the agency should deny the trademark.” GPT”.

Scher describes it this way: In the case of OpenAI, it would have to be established that “GPT” is proprietary and that the public perceives it as such instead of perceiving the acronym as referring to generative AI more broadly.

How would the USPTO rule on public perception? “One scenario is that you take a random sample of Americans and ask them to answer the question,” Scher says, but it’s a six-figure project that the government isn’t going to pay for, so any challenger of OpenAI should walk the bill for something like this.

Another way to establish public perception has to do with how “GPT” has been used in public, from late-night talk shows to public writing. “If people don’t treat it as proprietary, a trademark lawsuit would decide whether it’s copyrightable or not,” Scher says.

Unsurprisingly, this would involve a long process, which is surely the last thing OpenAI wants.

This begs the question of why the company didn’t act sooner to protect “GPT.” Here, Scher speculates that the company was “probably caught off guard” by its own success. (Indeed, it seems to be trying to get ahead of things in China, where it hasn’t launched ChatGPT yet and may not be allowed to, but it has reportedly attempted to register a related trademark.)

Either way, Scher says, it’s his opinion that “we’ve crossed a line where GPT isn’t three random letters. If a (startup) asked me if it was safe to adopt it, I would say that it is not safe.

OpenAI can further benefit from a facet of trademark law where the notoriety of a registered trademark is a dominant factor, Scher says. While you don’t have to be famous to get a mark, once an outfit is famous, it enjoys protection even outside of its sphere. Rolex is too famous a brand to be used on anything else, for example. If OpenAI can establish that “GPT” is a famous trademark, the company can also prevent the acronym from being used more broadly (although it would be costly to prosecute infringers).

This might be the only benefit for the company in this endless process. The more time passes and the more users OpenAI attracts and the more coverage the company receives, the more likely this last scenario becomes.

Is OpenAI known in the average household, by the average person? “Certainly,” Scher says, “they could come close to that.”