The inventions of the computer age are of a fundamentally different nature than the traditional mechanical, chemical, and manufacturing inventions of past centuries. The Patent and Trademark Office estimates that more than 15,000 applications for software patents are on file. Most of these inventions are essentially logical or mathematical in nature, involving "formulas" or "algorithms" that run applications, read / write / store data, compile code, display graphics, or calculate numbers. The mathematical nature of software inventions has caused courts and patent agencies to evaluate them with specialized treatment. The practicing programmer, engineer, or patent agent should be familiar with the legal boundaries within which mathematical and computational process inventions can be protected.
This paper focuses strictly on the statutory subject matter requirement for a utility patent. It does not discuss issues of novelty or nonobviousness, nor does it regard other forms of intellectual property protection such as copyrights.
In order to be patentable, an invention must satisfy four fundamental requirements. First, it must fall under the ambit of patentable "statutory subject matter" (SSM) as defined by 35 U.S.C. § 101. The invention must also be "useful," "novel," and "nonobvious." Most patent prosecution and litigation addresses the "nonobvious" requirement. Mathematical and computational processes, however, are an interesting special case. Challenges to the patentability of mathematical processes have generally focused on the § 101 SSM requirement.
Specifically, 35 U.S.C. § 101 defines SSM as any "process, machine, … , or … improvement thereof." As indicated in the 1952 Patent Act and popularized by the Supreme Court, this includes "everything under the sun that is made by man." Conversely, anything that is not "made by man" is not patentable. The Supreme Court has summarized the non-patentable subject matter as "laws of nature, natural phenomena, and abstract ideas." This makes sense, for the purpose of the patent system is to offer special protection to the inventions of an identifiable individual. The resources and properties of nature are supposed to be common property, available for everyone to share.
Mathematical / computational processes fall somewhere in the gray area between "processes" (patentable) and "laws of nature and abstract ideas" (unpatentable). For example, most people would agree with the Supreme Court's reasoning in Diamond v. Chakrabarty that "Einstein could not patent his celebrated law that E = mc^2; nor could Newton have patented the law of gravity. Such discoveries are 'manifestations of . . . nature, free to all men and reserved exclusively to none.'" What would the Court say about Newton's calculus? Mathematicians themselves are divided on the issue of whether mathematics is "invented" or "discovered." Some would regard calculus as a "law of nature". Others would characterize it as "made by man," but calculus would certainly fall under the "abstract idea" exception to patentability anyway.
When we consider more narrowly focused mathematical innovations, such as most modern computer code, we find ourselves crossing a very fuzzy line between the realms of "abstract idea" and "useful process." Today, as computer code is increasingly ubiquitous, specialized, and complex, the prevailing philosophy is that it is "invented" or "engineered" for particular purposes. Yet the jurisprudence on computer algorithms dates back to the dawn of the modern computer age. At that time, computation was not regarded as a useful end in and of itself, but as an "obvious" means toward automating or accelerating traditional practices. The principles of computation, moreover, were regarded as abstract mathematical rules akin to "laws of nature." The tension caused by the rapid shift in philosophy caused great confusion in the courts, and patent law lagged behind the realities of applied mathematics and computer science until just the last few years.
"In an earlier era, the PTO published guidelines essentially rejecting the notion that computer programs were patentable." The legal basis for patenting computational processes came with two Supreme Court decisions in the early 1980s. The first case was Chakrabarty, in which the Court not only gave SSM approval to "everything under the sun that is made by man," but also admonished the lower courts to "not read into the patent laws limitations and conditions which the legislature has not expressed." One year later, in Diamond v. Diehr, the Court expressly opened the door to the patentability of computer-assisted conventional processes. The critical language in Diehr is as follows:
When a claim containing a mathematical formula implements or applies that formula in a structure or process which, when considered as a whole, is performing a function which the patent laws were designed to protect … then the claim satisfies the requirements of § 101.
Where the invention or discovery is only of mathematics, the invention or discovery is not the "kind" of discovery the patent law was designed to protect and even the most narrowly drawn claim must fail.
This was the last word on the subject from the Supreme Court. The Court of Appeals for the Federal Circuit (CAFC), on the other hand, was highly active on this topic throughout the 1990s. The CAFC summarized the holding of Diehr thus: "mathematical algorithms are not patentable subject matter to the extent that they are merely abstract ideas." Starting from a similar premise, the CAFC in In re Alappat upheld the patentability of an entirely computerized process. The court explained, "this is not a disembodied mathematical concept which may be characterized as an 'abstract idea,' but rather a specific machine to produce a useful, concrete, and tangible result."
Alappat's "useful, concrete and tangible result" interpretation of Diehr is still the applicable legal test of the SSM requirement for computational processes.
In practice, the words "concrete and tangible" have not had much of a bite. Since the Alappat test was formulated in 1994, no federal appeals court has struck down a computational process patent on the basis of its lacking a "concrete and tangible result." Meanwhile, many processes with "intangible" results, even purely numerical results, have passed muster under the Alappat test. The CAFC explained this standard in AT&T v. Excel in 1999: "Since the process of manipulation of numbers is a fundamental part of computer technology, we have had to reexamine the rules that govern the patentability of such technology." In other words, a process that merely affects or makes use of a computer is regarded as sufficiently "concrete and tangible" to qualify as patentable SSM, provided that it is "useful."
Without a computer to implement it, a raw mathematical formula is still regarded as non-SSM, either because it is too abstract or because it does not produce a "concrete and tangible result." However, it is difficult to conceive of a mathematical algorithm that could not be implemented on a computer.
A patented invention must also be "useful;" software must be designed with some particular end use in mind. Software that merely crunches numbers or generates a curve with no application, that flips switches at random, or that proves theorems with no known relevance outside of pure mathematics or logic, would fail to meet the usefulness test. A program whose sole purpose was to spread a catastrophic virus would probably be struck down under usefulness arguments as well, unless perhaps there were conceivable military applications.
A few tests for patentability have come and gone, and some "conventional wisdom" on the subject is not supported by actual legal doctrine.
"You can't patent an 'algorithm'"
Not true. An algorithm, even a mathematical algorithm, is patentable unless (1) it is not sufficiently disclosed to overcome the "abstract idea" exception, or (2) it is executed with no useful application in mind. This description of just what an "algorithm" is may be helpful: "Although one may devise a computer algorithm for the Pythagorean theorem, it is the step-by-step process which instructs the computer to solve the theorem which is the algorithm, rather than the theorem itself." Under Alappat, the algorithm might be patentable. The theorem itself, being an abstract idea, certainly would not.
"You can't patent a 'business method.'"
False. This was a long-standing unfounded rumor based on a misreading of case law. Neither the Supreme Court nor the CAFC ever struck down a method patent strictly on the basis that it was useful for business. In State St., the CAFC said definitively, "We take this opportunity to lay this ill-conceived exception to rest."
"You can't patent a process unless it involves a 'physical' element.'"
False. The Diehr court lists "transforming or reducing an article to a different state or thing" as "an example, not an exclusive requirement" of patentability. The Court of Custom and Patent Appeals (CCPA)'s Freeman-Walter-Abele test, which required a mathematical algorithm to be "applied to or limited by physical elements," has been seriously called into question by Alappat, State St., and AT&T. "Whatever may be left of the earlier test, if anything, this type of physical limitations analysis seems of little value." As discussed above, patents are being issued and upheld on inventions that amount to no more than calculating a number, as long as that number is useful information.
"After Diehr and Alappat, the mere fact that a claimed invention involves inputting numbers, calculating numbers, outputting numbers, and storing numbers, in and of itself, would not render it nonstatutory subject matter, unless, of course, its operation does not produce a 'useful, concrete and tangible result.'"
"Abstract ideas" are still unpatentable. If a process is entirely computational, it can be patented only if the algorithm is fully disclosed, and it must ultimately serve some particular application for human users.
