Abstract
In recent years, our federal courts have given increased attention to the question of what subject matter is eligible for patent protection. The resulting caselaw, developed mostly in the context of business methods or other relatively straight forward technologies, exhibits a number of trends that broadly call into question the patentability of inventions in the field of artificial intelligence. In particular, one series of cases has revitalized the “mental steps doctrine” as a mechanism for invalidating patents. These cases suggest that technology for emulating or replicating activities that could otherwise be accomplished by the human thought process are not patentable. A second series of cases has placed undue emphasis on quantifiable operational improvements as a yardstick for patent-eligibility of computer-related inventions. These cases suggest that even the most ingenious and useful advances in that area may be unpatentable if they do not also provide a readily measurable improvement in performance. Although this precedent was largely crafted outside the artificial intelligence context, it is nonetheless being used by inventors, investors, and courts to gauge the patentability of advances in artificial intelligence. As a result, incentives to innovate in that field are being considerably diminished and, in some instances, altogether eliminated—a consequence that does not appear to have been considered by the deciding courts. This Article highlights this growing problem, explains why the eligibility barriers developed in the series of cases described above should generally not be applied to prevent patenting of advances in artificial intelligence, and proposes better ways forward.