Artificial Intelligence Inventions & Patent Disclosure

Artificial intelligence (“AI”) has attracted significant attention and has imposed challenges for society. Yet surprisingly, scholars have paid little attention to the impediments AI imposes on patent law’s disclosure function from the lenses of theory and policy. Patents are conditioned on inventors describing their inventions, but the inner workings and the use of AI in the inventive process are not properly understood or are largely unknown. The lack of transparency of the parameters of the AI inventive process or the use of AI makes it difficult to enable a future use of AI to achieve the same end state. While patent law’s enablement doctrine focuses on the particular result of the invention process, in contrast, this Article suggests that AI presents a lack of transparency and difficulty in replication that profoundly and fundamentally challenge disclosure theory in patent law. A reasonable onlooker or a patent examiner may find it difficult to explain the inner workings of AI. But even more pressing is a non-detection problem—an overall lack of disclosure of unidentified AI inventions, or knowing whether the particular end state was produced by the use of AI.The complexities of AI require enhancing the disclosure requirement since the peculiar characteristics of the end state cannot be described by the inventive process that produced it. This Article introduces a taxonomy of AI and argues that an enhanced AI patent disclosure requirement mitigates concerns surrounding the explainability of AI-based tools and the inherent inscrutability of AI-generated output. Such emphasis of patent disclosure for AI may steer some inventors toward trade secrecy and push others to seek patent protection against would-be patent infringers despite added ex ante costs and efforts. Utilitarian and Lockean theories suggest justifications for enhanced AI patent disclosure while recognizing some objections. Turning to the prescriptive, this Article proposes and assesses, as means for achieving enhanced disclosure, a variety of disclosure-specific incentives and data deposits for AI. It concludes by offering insights for innovation and for a future empirical study to verify its theoretical underpinnings