USPTO’s Approach to AI and Machine Learning Patents Hinges on ‘Abstract Ideas’ Exception

Answer first: The United States Patent and Trademark Office (USPTO) has issued limited explicit guidance on patenting artificial intelligence (AI) and machine learning (ML) inventions, relying largely on existing frameworks for software patents. Central to patent eligibility is the ‘abstract ideas’ exception, which has been shaped…

Patents context for IP teams

The United States Patent and Trademark Office (USPTO) has taken a relatively reserved stance in issuing explicit guidance specifically tailored to machine learning (ML) and artificial intelligence (AI) patents, especially when compared to the European Patent Office (EPO). Instead, the USPTO’s approach largely builds on the foundational criteria for patent eligibility, focusing on whether the subject matter qualifies as patentable under statutory requirements.

Patent eligibility under US law requires that an invention be useful and not fall within certain statutory exceptions. Among these exceptions, the ‘abstract ideas’ category is particularly pertinent to AI and ML inventions. This exception was notably clarified in the landmark 2014 Supreme Court decision Alice Corp. v. CLS Bank International, which set the precedent for modern interpretations of software patent eligibility at the USPTO.

Key takeaways for USPTO AI patent eligibility

• Confirm how the development affects patents ownership, enforcement, licensing, or portfolio records.
• Separate confirmed facts from legal interpretation before advising business teams.
• Map deadlines, affected assets, contracts, and evidence files to the responsible internal owner.
• Use the issue as a prompt for monitoring, filing strategy, dispute preparation, or member education.

Practical analysis

A practical illustration of the ‘abstract ideas’ exception is found in the case Purepredictive Inc v H2O AI Inc, where a California district court ruled that a machine learning invention applying purely to data was too abstract to merit patent protection. The court reasoned that the invention merely used computers as tools without providing a specific improvement to computer technology itself. The claimed invention involved a three-step process of learning mathematical functions, evaluating them, and selecting the best one, which the court deemed insufficiently concrete.

In response to ongoing uncertainty, the USPTO released guidelines in January 2019 aimed at clarifying the ‘abstract ideas’ exception within the context of software inventions. Although these guidelines are not AI- or ML-specific, a 2020 USPTO report on a 2019 Request for Comment project indicated that most respondents viewed AI as a subset of computer-implemented inventions. Consequently, the existing USPTO guidance was considered adequate to address AI-related patent applications.

The 2019 guidelines categorize abstract ideas into three groups, with particular focus on the ‘mental processes’ exception. The USPTO has provided hypothetical examples illustrating how claims involving mental processes, such as ranking items, are ineligible for patent protection if the process could be performed by a human without a computer. This underscores that merely implementing a mental process on a computer does not satisfy the requirement for patent eligibility.

Conversely, inventions that produce tangible outputs—such as manipulating images or generating sound files—are more likely to qualify for patent protection, as these outputs cannot be replicated purely mentally. Legal experts emphasize that to be patent-eligible, an algorithm-based invention must advance a specific technical application and not simply solve a problem through abstract calculation. Patent applications must detail how the algorithm interacts with physical computer infrastructure and address a real-world problem.

Even when a claim involves a mental process, patent examiners are instructed to determine whether the process is “integrated into a practical application.” This means that the claim as a whole must demonstrate practical implementation beyond abstract mental steps. Simply adding a computer to perform the process does not suffice to establish patent eligibility.

IP Watchdog highlights the importance of clearly defining the innovation, the problem it solves, and the technical details of how it achieves its function. Applications lacking this clarity risk rejection.

- A method and apparatus for performing dynamic textual complexity analysis using machine learning AI, which, despite involving a process that could be performed mentally, provides predictive outputs that satisfy practical application requirements.

- A system and method for machine learning predictive maintenance through auditory detection on natural gas compressors, which addresses a concrete technical problem by classifying auditory signals to detect anomalies.

- A method for selecting, analyzing, and summarizing content objects based on user requests, where the clearly defined output (the summary) supports patent eligibility despite involving mental calculation elements.

These examples illustrate that AI and ML inventions with defined technical applications and tangible outputs are more likely to overcome the ‘abstract ideas’ hurdle at the USPTO. The office’s current stance encourages applicants to focus on practical implementations and detailed descriptions of the technical problems their inventions solve.

Related IIPLA reading

• Latest IP news
• IP resources library
• IP guide
• IIPLA events and webinars