OpenAI’s patent licensing “promise” is not what it seems

Generative artificial intelligence (genAI) company OpenAI recently published its “approach to patents,” which includes what might appear to be a promise not to assert its own patents against other parties except in self-defense. It is less of a promise than it seems.

The statement is short, and the most substantive portion is its last sentence, “We pledge to only use our patents defensively, so long as a party does not threaten or assert a claim, initiate a proceeding, help someone else in such activities against us, or engage in activities that harm us or our users.” The end of that sentence carves out so many broad exceptions to the defensive-use-only pledge that the statement as a whole ends up being nothing more than a paper tiger – public relations virtue – signaling to the tech community and regulators rather than a substantive attempt to foster healthy competition in the marketplace.  

Consider just the last exception, “so long as a party does not…engage in activities that harm us or our users.” This statement can be reasonably construed to mean that OpenAI may sue for patent infringement if an entity competes with OpenAI in any market or even levels criticism (regardless of whether it is fair and accurate or not) against OpenAI or its customers.[1] Additionally, if an entity provides a product or service that generates sub-par output to customers it has in common with OpenAI – e.g., a genAI model that hallucinates – OpenAI could allege harm to its users and sue that entity for patent infringement. 

But more fundamentally, a statement on a website does not have the same legal effect as a contract signed by parties who are bound to its terms, where something of value has been exchanged between these parties. It is an aspiration, not an obligation. Thus, the statement would not constitute an affirmative defense to an accusation of patent infringement made by OpenAI. Also, OpenAI can unilaterally change its terms at any point or rescind it altogether. And if OpenAI sells or licenses the rights to its patents to a third party, it is unlikely that this non-assertion pledge will travel with the assets.  

Further, context matters. Thanks to misapprehension by Congress and the federal courts regarding so-called “patent trolls” (and significant lobbying efforts by Big Tech), patent rights have been considerably weakened over the last two decades. In particular, software patents are difficult to obtain and even more difficult to enforce. And because of the expense and burden of infringement litigation, companies generally pursue cases only when an infringer harms their finances or goodwill. Again, the exception for parties who “engage in activities that harm” OpenAI or its users would come into play and could be read broadly enough to include anyone who undercuts OpenAI’s market position. Therefore, even if this broader promise not to assert patent rights were binding, OpenAI is not giving up much – if anything – by making it.  

Just as importantly, OpenAI’s statement relates only to enforcement of patents; OpenAI has not chosen patenting as its primary IP protection strategy. In comparison to other Big Tech companies, who obtain thousands of patents every year, a quick search shows that OpenAI has only 11 granted U.S. patents and has just a handful of published applications.[2] In comparison, OpenAI’s partner Microsoft was granted over 1,800 patents in 2023 alone. Instead, OpenAI’s current generation of genAI models appear to be protected as closely held trade secrets that have not been patented. Thus, even technologically, OpenAI’s promise does not cover much. 

A decade ago, Tesla CEO Elon Musk made a similar promise not to assert the company’s patents against “anyone who, in good faith, wants to use our technology.” It was an empty promise then, and OpenAI’s position has even more loopholes and exceptions.  

One should not be fooled by the AI hype coming out of Silicon Valley – while genAI tools can be remarkably effective at a number of specific tasks, the rush to market and sell undertested models greatly reduces their practical use. Likewise, one should not be fooled by an AI company making unenforceable non-commitments about other parties being able to freely use its patented technology. In short, at a time when OpenAI is being criticized heavily and losing key staffers, it appears to be seeking to curry favor by making an announcement that is less than meets the eye. 

[1] One might even question whether writing a blog post criticizing OpenAI’s “approach to patents” would be considered harmful to OpenAI.
[2] Due to the U.S. Patent and Trademark Office’s 18-month publication delay as well as an applicant’s ability to file patent applications with non-publication requests, OpenAI may have more pending applications that cannot currently be viewed by the public.