The United States Patent and Trademark Office (USPTO) has announced a new initiative to explore the use of artificial intelligence in enhancing its patent and trademark examination processes. In a recently issued Request for Information (RFI), the agency is seeking input from AI vendors on tools that could streamline the search and analysis of prior art. Specifically, the USPTO is interested in AI systems capable of analyzing both domestic and international patents as well as non-patent literature, evaluating the relevance of this information to new patent applications, and identifying overlapping material to assess novelty and non-obviousness.
However, there are significant limitations and concerns tied to this opportunity. Notably, the USPTO makes it clear that selected vendors should expect "primarily non-monetary" compensation offering exposure, visibility, and the prestige of contributing to a federal initiative in lieu of direct payment. In addition, any chosen vendors must be able to provide their own infrastructure including compute power, storage, and networking within the USPTO’s secure cloud environment.
A critical issue raised by this initiative is the risk of AI "hallucinations," where AI tools may misidentify or incorrectly cite legal and patent documents. Such errors could severely impact the integrity of patent evaluations. While the USPTO acknowledges that human oversight will remain essential, the RFI does not detail how this oversight will be implemented or enforced.
In summary, the USPTO is in the early stages of investigating how AI might improve the efficiency and consistency of patent examination. While no contracts have been awarded, the agency is inviting vendors to contribute innovative solutions offering recognition rather than financial compensation. Human supervision will remain a cornerstone of the process to mitigate the risks inherent in AI-driven analysis.
Read the article here
