U.S. District Court Holds that AI Algorithms Cannot Be Listed as Inventors on Patents

As we discussed earlier, a variety of artificial intelligence (AI) technologies exist and are being used in biopharma (e.g., discovery and development of drugs, optimization of clinical trial design, identification of novel biological targets and pathways of interest, analysis and use of biomedical, clinical, and patient data, and many others). As in most industries, AI will play an increasingly larger role in biopharma patents. However, there are a number of challenges of AI in biopharma patents. One of these challenges that has recently come to the forefront is the question of inventorship.

Patent systems around the globe offer a quid pro quo that exchanges limited monopolies for disclosures of inventions. Most patent filings list: (1) the inventor(s); and (2) the applicant. The applicant may be an assignee (e.g., company, university, organization, etc.) with rights to seek patent protection on innovations identified during employment and within the scope of employment. Frequently, the assignee is a current or former employer of the inventor(s).

In some jurisdictions (e.g., in the United States), the inventor(s) hold the rights to prosecute the patent application and assert any resulting patent unless and until the inventor(s) assign those rights to another entity (which is frequently done in employment and work for hire contracts). In some jurisdictions (e.g., the European Union), it is presumed that the party that applied for a patent holds the rights to the patent application.

Thus, it is well-established that non-human entities may be the applicant, assignee, and/or owner of a patent. However, it is not well-established that a non-human entity may be an inventor on a patent application. Multiple patent offices (e.g., USPTO, UKIPO, and WIPO) have been considering what the standard in this respect should be. A recent U.S. district court case addressed the issue of whether a non-human entity may be an inventor and found that they could not. The case, which is discussed below, will be appealed to the U.S. Court of Appeals for the Federal Circuit, and, as there was before the district court, there will be an opportunity for interested parties to submit amicus briefs on this issue. 

On September 2, the U.S. District Court for the Eastern District of Virginia issued a decision granting a Motion for Summary Judgment for the United States Patent and Trademark Office (USPTO) and upholding the Office’s view that AI algorithms cannot be listed as inventors on U.S. patents. The court pointed to the Administrative Procedures Act’s (APA’s) strong deference to final agency decisions, barring any egregious errors.

DABUS generated outputs corresponding to (1) a fractal design for food container surfaces that may help prevent stacked containers from sticking together and (2) a technique for controlling the timing of flashing warning lights to help attract attention. Dr. Stephen Thaler (DABUS’s creator and owner) filed patent applications on these inventions that were filed around the world, listing Thaler as the applicant and listing only DABUS as the inventor.

Mixed Success

Thaler has had mixed success in obtaining patent protection for DABUS’s inventions across the globe. In the United Kingdom and Europe, the UK Intellectual Property Office (UKIPO) and European Patent Office (EPO) rejected the idea that an artificially intelligent machine qualified as an inventor under their patent laws. More recently,  however, South Africa granted Thaler’s patent applications and became the first nation to grant a patent to an AI inventor (notably, South Africa has a registration-based patent system, where a patent is granted so long as all relevant forms are submitted and correctly completed). Australia’s patent office, which substantively examines patent applications, originally rejected Thaler’s patent application based on the fact that only a human could be named as an inventor on an Australian patent application. However, a federal court overturned that decision, thereby indicating that listing an AI algorithm as an inventor is permissible (the Australian Commissioner of Patents has since appealed).

In the United States, the USPTO determined that listing an AI algorithm as an inventor was not permitted, and the agency thus declined to examine the patent applications on their merits. Thaler and his legal team (led by Dr. Ryan Abbott) challenged this decision by filing suit in the Eastern District of Virginia.

U.S. Court: USPTO Deserves Deference

The district court’s September 2 decision is based foremost on the deference granted to an administrative agency interpreting its own laws, and on the standard of review. The decision begins by stating that under the APA, a final agency decision may only be set aside if it is arbitrary, capricious, or otherwise not in accordance with the law, and then states that an administrative agency is accorded deference in its carefully considered interpretation of statutory provisions.

Thaler had argued that that the USPTO was not due deference because it allegedly did not consider alternative interpretations or provide evidence of Congressional intent. The judge rejected these for two reasons: first, it found that Thaler’s arguments added requirements that were counter to Supreme Court and Federal Circuit precedent and second, it determined that the USPTO’s decision was carefully considered.

The decision goes on to state that, aside from the issues of deference and standard of review, the USPTO correctly decided that an “inventor” must be a human being. The court bases this conclusion on an analysis of Congress’ definition of “inventor” and “joint inventor” in both the Patent Act and the more recent 2011 America Invents Act as an individual. The decision then examines a Supreme Court ruling interpreting the Torture Victim Protection Act, multiple dictionaries, and common usage to find that “individual” referred to a “natural person,” and therefore that an inventor must be a natural person. The decision concludes by citing a number of Federal Circuit decisions stating that inventors must be natural persons and finding that policy considerations such as those raised by Thaler could not overcome a statute’s plain language.

When the Time Comes, it’s Up to Congress

The decision ends with the interesting assertion that artificial intelligence may someday reach a level of sophistication sufficient to establish inventorship, but that when that time comes it will be up to Congress, not the courts, to decide whether or not to expand the definition of inventor to include it. It will be interesting to see if future courts or the USPTO will similarly defer to Congress on this issue.

Thaler has already indicated that he intends to appeal the summary judgment determination. Thus, interested parties will, once again, have an opportunity to file an amicus brief (this time, to the Federal Circuit) to present considerations pertaining to the issue.

Knowledge assets are defined in the study as confidential information critical to the development, performance and marketing of a company’s core business, other than personal information that would trigger notice requirements under law. For example,
The new study shows dramatic increases in threats and awareness of threats to these “crown jewels,” as well as dramatic improvements in addressing those threats by the highest performing organizations. Awareness of the risk to knowledge assets increased as more respondents acknowledged that their