Major technology paradigm shifts are typically followed by a wave of patent application filings. If you don’t believe that, a quick patent search will validate this for your technology of choice.

What you’ll also find is that there can be a significant lag between when the technology enters the mainstream and when related patent applications actually publish. Part of this is driven by the reality that inventions can’t be generated out of thin air, and it takes real time and resources to develop the kinds of technical solutions to technical problems that underlie strong patent applications.

Another factor is the ~18 month publication lag between when most patent applications are filed to when they publish. This means that even if a technology has widespread adoption, inventions are being developed, and patent applications filed as soon as the technology emerges, there will be no publicly available evidence of those filings until after the publication lag.

Generative AI and LLMs

Now we come to recent advances in generative artificial intelligence (AI) and large language models (LLM). Considering the start date of this wave as ChatGPT 3.5’s November 30, 2022 launch, we should expect that the number of published patent applications referencing the term “ChatGPT” will be fairly low until at least June 2024. Indeed, as of today (January 25, 2024) there are only about 60 U.S. patent filings that use the term. Nevertheless, there are likely to be a vast number of filings that are still hidden by the publication lag (i.e., filed between December 2022 and today), but will start becoming available in the second half of 2024 and beyond.

What does this mean for companies developing patent filing strategies? There is a small window of opportunity over the next six months or so to get applications on file – and into examination – before the wave hits. The amount of prior art available to reject applications where the inventive concepts focus on LLM technologies (and likely other GAI implementations using diffusion models, etc.) will increase as the wave hits, resulting in longer, more expensive examination.

There’s an important caveat to these considerations, due to the first to file rule for prior art: the currently hidden patent applications will eventually publish and could be used as prior art against applications first filed during the publication lag but examined after those hidden applications publish. This could come up with continuation filings as well as invalidity challenges in litigation or inter partes reviews.

Nevertheless, having at least the first patent application in a family examined under favorable conditions can improve the overall success of the family. This is because there is a path dependency to examination and portfolio development: patent examination is most effective as a collaborative process with the examiner. Working with the examiner to identify patentable subject matter early pays real dividends by leading to efficient examination in each subsequent filing, saving overall costs while mitigating prosecution history estoppel and other patent examination risks. In addition, a secondary effect of less prior art relating to LLMs being available in 2024 is that overcoming 101 abstract idea rejections may be easier today than in 1-2 years when the perception of how inventive using LLMs is has changed.