OpenAI Wins Trademark Injunction

A California federal court has ruled that a Silicon Valley entrepreneur may not use the trademark “OpenAI” because of a likelihood of confusion with Microsoft’s OpenAI.

OpenAI, the founder of the generative artificial intelligence (AI) tool ChatGPT, among other things, brought a trademark infringement suit against defendant Guy Ravine and his company Open Artificial Intelligence.

The plaintiff uses the trademark OpenAI (no space), and the defendants use Open AI (with a space).

The parties disagree about who the senior user of the mark is.

The plaintiff sought a preliminary injunction to prohibit the defendants from using the mark.

As the court noted,

Plaintiff’s trademark is associated with perhaps the most renowned artificial intelligence tools currently in use and, quite quickly, became ubiquitous. Defendants’ trademark, by contrast, is associated with a website that, until shortly before this litigation commenced, was inoperable at best. Only plaintiff, therefore, has shown that it has a protectable interest in its trademark.

As the court discussed, plaintiff OpenAI announced its foundation in December 2015 and has since introduced several very popular AI products. It uses the OpenAI mark in association with its goods and services, including its website, social media, and marketing. In the month before it filed its complaint, the plaintiff’s site had 185 million US visitors.

Defendant Ravine developed Wikineering.org, a website that offered articles related to engineering topics, including AI, in 2012. He claims that in 2015, he branded a portion of the site with the Open AI mark on a page called the “Initial Collaboration Tool.” He says this page was available until early 2016, but no record of it exists, and Ravine was unable to submit anything to the court to corroborate its existence.

Ravine submitted a mockup of how he claimed he remembered the page to look. However, the plaintiff’s expert showed that the supposed content for the page is “virtually identical to content in a 2017 college textbook,”—which was not published until two years after defendant Ravine said the content appeared on his platform.

Open.AI

In March of 2015, Ravine bought the domain name Open.ai. The website consisted of a landing page with the term Open AI, calling it an “Open Industry Wide & Academia Wide Deep Learning Initiative.” The site was inoperative until January 24, 2022. On that date, Ravine added a third-party plugin for a text-to-image generator called Stable Diffusion, a tool that competes with the plaintiff’s Dall-E 2 text-to-image generator.

On December 11, 2015, the night after the press reported the plaintiff’s founding, Ravine filed an application to register the Open AI mark in the Principal Register, claiming he had used the mark in commerce since at least March 25, 2015.

Ravine also wrote to the plaintiff’s CEO and co-founder, Sam Altman, and told him that he had also been working on an AI project. Plaintiff’s president and co-founder, Greg Brockman, met with Ravine soon after but rejected his suggestion that the parties collaborate on AI development. Brockman did offer to purchase the Open.ai domain if Ravine changed the name of his upcoming project, but Ravine declined.

Reject Trademark Application

In March 2016, the USPTO rejected Ravine’s trademark application, finding that the specimen he submitted on his placeholder website didn’t show that he’d used the mark in commerce.

The USPTO determined that Ravine’s mark was “merely descriptive” and thus ineligible for the Principal Register. Four months later, Ravine registered the mark on the Supplemental Register.

Shortly before the lawsuit was filed, Ravine redesigned his website to look like the plaintiff’s.

The plaintiff first tried to register its trademark with the US Patent and Trademark Office (USPTO) in September 2016.

The USPTO acknowledged that there were no similar “registered” marks but warned the plaintiff that defendant Ravine’s prior-filed application “may present a bar to registration” because “of a likelihood of confusion.”

Also, the USPTO noted that the mark OpenAI was “merely descriptive.”

The plaintiff didn’t proceed with the registration, and the USPTO closed the application as abandoned in August 2017.

Years later, in January 2022, the plaintiff once again tried to register the OpenAI mark, and once again, the USPTO rejected it as “merely descriptive.” This time, the plaintiff submitted evidence that the mark had “acquired distinctiveness” and thus should be registered.

The USPTO rejected this claim, noting that although the plaintiff had used the mark for five years, this was “insufficient to show acquired distinctiveness because the applied-for mark is highly descriptive of applicant’s goods and/or services.”

After Ravine sent the USPTO a letter of protest in which he argued the plaintiff’s mark was confusingly similar to his own, the USPTO provisionally rejected the plaintiff’s application.

In February 2022, Altman again asked Ravine about buying the Open.ai domain and related intellectual property (IP).

Ravine responded:

Elon Musk paid $11 million for the Tesla domain and trademark in 2017. As we both know, OpenAI has the potential to become larger than Tesla and, in either event, will become one of the largest companies in the world in a relatively short period.

However, Ravine said he had “no use for the money. As an individual, I’m already well off.” Instead, he suggested that the plaintiff donate the purchase price to an academic collaboration.

The plaintiff never did this.

As the court described,

The parties agree that (i) the defendants have a registered mark on the USPTO’s Supplemental Register and (ii) both parties’ trademark applications on the Principal Registry were denied. Regardless, the plaintiff argues that it is the senior user because it was the first one to actually use the mark in the provision of goods or services. Defendants response that they have used the mark in commerce since March 2015. Moreover, defendants argue that plaintiff had ample notice of their use because, among other things, the USPTO rejected plaintiff’s trademark application as confusingly similar to defendants’ mark after the USPTO published defendants’ trademark on the Supplemental Register. Defendants assert that this prior use is dispositive.

The court noted that the core issue is whether either entity had a common law right to the trademark:

To acquire ownership of a trademark it is not enough to have invented the mark first or even to have registered it first; the party claiming ownership must have been the first to actually use the mark in the sale of goods or services. Therefore, a party pursuing a trademark claim must meet a threshold ‘use in commerce’ requirement.

Here, said the court, plaintiff

has demonstrated that its users associate its trademark with its products; its extensive use of the trademark in advertising its products; and its use of the trademark since December 2015, which in the past year has made its trademark one of the most recognized in the artificial intelligence industry, if not the world.

Also, said the court, “the evidence strongly suggests that defendants registered their mark merely to reserve a right to it rather than for a genuine, commercial purpose.”

The Court found that the balance of equities and public interest favored a preliminary injunction against the defendant’s use of the mark.