On October 6, 2020, the United States Patent and Trademark office (USPTO) released a report titled “Public Views on Artificial Intelligence and Intellectual property Policy.” The report follows the USPTO’s August 2019 request for comments on patenting AI inventions, and the USPTO’s October 2019 request for comments related to the impact of AI on other IP policy areas.
A few general themes emerged from the nearly 200 submissions, including “the fact that AI has no universally recognized definition,” “the current state of the art is limited to ‘narrow’ AI”, as opposed to artificial general intelligence akin to human intelligence, and “a general sense that existing U.S. intellectual property laws are calibrated correctly to address the evolution of AI.” Report pp. ii-iii.
With respect to patent protection, the consensus remains that inventors must be human, but some differences arose as to what human activities should qualify as a contribution to the conception of an invention. The Report notes that “activities such as designing the architecture of the AI system, choosing the specific data to provide to the AI system, developing the algorithm to permit the AI system to process that data, and other activities not expressly listed here may be adequate to qualify as a contribution to the conception of the invention.” Report, p. 5. But perhaps the USPTO will need to revisit the question of whether machines can be inventors when and if science agrees that machines can “think” on their own.
On data and new forms of IP protection, the Report acknowledges that data “is a foundational component of AI,” and “data and datasets, including their collection and compiling, have value.” Report p. 15. Commentators, however, were divided between the view that new IP rights were necessary and the belief that the current US IP framework was adequate to address AI inventions. No concrete proposals were submitted on how any newly created IP right for AI and data should function, and many called upon the USPTO to further consult the public on the issue.
On the copyright front, the Report notes that existing “statutory and case law should adequately address the legality of machine ‘ingestion’ in AI scenarios,” and that mass digitization and text and data mining “may be considered copyright infringement or fair use, depending on the facts and circumstances at issue.” Report p. 23. Critically, however, “mass digitization for purposes of machine learning (ML) ‘ingestion’ processes-and large-scale ingestion of already-digitized works-has not yet been tested by the courts.” Id. p. 25. Additionally, commentators mostly found that existing laws are adequate to continue to protect AI-related databases and datasets—including copyright protection as a compilation, trade secret, contract law, and tort law—and there is no need for a sui generis database protection law, such as exists in Europe.
In view of these comments and the areas of greatest divergence, there are a few practical takeaways to keep in mind. First, there are numerous ways a human can contribute to the conception of an AI invention—from curation of the input data to practical application of the output—and innovators should document and evaluate who was involved at every step of the AI development process as part of the inventorship inquiry for patent protection. Second, datasets used for AI can have tremendous value, and there are multiple ways to preserve the value of datasets, such as with copyright, trade secret, and contractual/licensing protection. As each form of protection has its strengths, it is important to choose the right avenues of protection for the right datasets and applications. Finally, we will likely see further developments in the case law on the issue of copyright infringement through machine ingestion of data, and the applicability (or not) of the fair use defense. The Oracle v. Google Supreme Court case on fair use may potentially touch these issues, as well as the numerous data scraping cases making their way through the courts.