Earlier this year, the Intellectual Property Owners Association (IPO) released its White Paper on Best Practices for Protecting Inventions Relating to Artificial Intelligence (White Paper). The extensive guidance discusses the patenting process concerning AI in the United States as well as in other jurisdictions and aims to aid practitioners and applicants in forming sound decisions.

Key takeaways from the guidance

The document begins with recommendations for preparing and conducting a disclosure interview with an AI inventor. In particular, sample questions regarding the invention, such as technical challenges, solutions, advantages and inputs, should allow practitioners to achieve strong patent protection in favor of the AI inventors.

The paper then goes on to set out the ethical issues that should be considered when preparing the invention disclosures. According to the IPO, practitioners need to evaluate aspects that could cause a statistical bias and thus a disproportionate impact on people by discriminating users, violating privacy or being exploited by malicious customers. Practitioners therefore are advised to use a “garbage in, garbage out” approach, i.e., to conclude that if the data used to build an AI model is faulty, one can expect that the AI model will produce a faulty output. Another recommendation is to connect the inventor with an AI ethics partner.

Further topics concern the drafting of claims in order to avoid prosecution of AI inventions under Title 35 US Code § 101 and 112. The paper also points out the distinction between features that need to be made detectable, i.e., that provide a basis for infringement allegations, and features which can be left out from disclosure and thus maintained as trade secrets. Finally, the paper analyzes patent litigation cases before the US courts concerning the enforcement of AI and other software-related technologies.

US vs. the world: Proving patent eligibility as the key difference to other jurisdictions

One of the major challenges for inventors and practitioners is establishing patent eligibility. As evident from the White Paper, this requirement also causes the biggest differences between the various patent offices.

The European Patent Office (EPO), for instance, has deemed AI-related inventions abstract per se—and hence not patent-eligible, as they are based on computational models and algorithms. According to the EPO Guidelines, a patent may be granted only when the AI can be applied so as to solve a specific technical problem.

On the other hand, the United States Patent and Trademark Office does not consider AI-based inventions as abstract per se. Rather, the office examines whether AI-related claims include an abstract idea (e.g., a method of organizing human activity, a mathematical relationship, etc.). If that is the case, authorities and courts treat the claims accordingly by considering whether they are integrated into a practical application and whether additional elements amount to “significantly more” than the abstract idea (the so-called “Alice/Mayo” test).

China also lays down particular rules for the examination of inventions that include algorithms, business rules, or method features. Among others, the claim as a whole should be considered while it should be clear how the technical features contribute to the solution of a technical problem.

Contrastingly, Korea as well as other countries appear to be less stringent towards the requirement of patent eligibility, as they tend to put the focus on the novelty and inventiveness of the invention in suit.

Outlook: Australia and South Africa as AI protection havens?

Despite their differences concerning the regulation of certain aspects of the patentability of AI-related inventions, all IP5 members (comprising the patent offices of the United States, Europe, Japan, Korea and China) share the same opinion on patent ownership. As the White Paper states, all those countries have refused to approve patents in which the AI itself is named as the inventor. The current wording of the statutes (e.g. “Whoever”, “any natural or legal person”) thereby serves as one of the strongest arguments against the entitlement of AIs to a patent.

On the other side of the spectrum, however, are Australia and South Africa. Both countries have confirmed that AI tools may also be listed as the inventors of AI-based creations. As this decision could equip researchers and AI developers with wider discretion in the patenting process, two new innovation powerhouses might have been born.

Nevertheless, in view of the concrete examples, pinpoints and guidance included in the IPO White Paper, the role of the US as one of the leading forces in AI research and development does not seem endangered. Instead, the paper serves as a clear sign that inventors and practitioners can rely on a comprehensive framework that attempts to efficiently protect the fruits of one’s labor.