Why a 40-year-old SCOTUS ruling against software patents still matters today

Forty years ago this week, in the case of Parker v. Flook, the US Supreme Court came close to banning software patents. “The court said, ‘Well, software is just math; you can’t patent math,’” said Stanford legal scholar Mark Lemley. As a result, “It was close to impossible in the 1970s to get software patents.”

If the courts had faithfully applied the principles behind the Flook ruling over the last 40 years, there would be far fewer software patents on the books today. But that’s not how things turned out. By 2000, other US courts had dismantled meaningful limits on patenting software—a situation exemplified by Amazon’s infamous 1999 patent on the concept of shopping with one click. Software patents proliferated, and patent trolls became a serious problem.

But the pendulum eventually swung the other way. A landmark 2014 Supreme Court decision called CLS Bank v. Alice—which also marks its anniversary this week—set off an earthquake in the software patent world. In the first three years after Alice, the Federal Circuit Court, which hears all patent law appeals, rejected 92.3 percent of the patents challenged under the Alice precedent.

The shifting rules about software patentability reflect a long-running tug of war between the Supreme Court and the Federal Circuit. The Federal Circuit loves software patents; the Supreme Court is more skeptical.

That fight continues today. While the Federal Circuit has invalidated many software patents in the four years since the Alice ruling, it also seems to be looking for legal theories that could justify more software patents. Only continued vigilance from the Supreme Court is likely to ensure things don’t get out of hand again.

The 40-year-old Flook ruling remains a key weapon in the Supreme Court’s arsenal. It’s the court’s strongest statement against patenting software. And, while software patent supporters aren’t happy about it, it’s still the law of the land.

The banhammer?

In the mid-1970s, Dale Flook applied for a patent on the concept of using a computer to monitor a process of catalytic conversion of hydrocarbons. Software would trigger an alarm if the temperature, pressure, or other variables got out of control.

“Fixed alarm limits may be appropriate for a steady operation, but during transient operating situations, such as startup, it may be necessary to update the alarm limits periodically,” the Supreme Court explained in its 1978 ruling. Flook’s patent provided a specific formula for updating the “alarm limit” that determined when an alarm would sound.

But the Supreme Court concluded that Flook’s invention was not eligible for a patent. “The patent application does not purport to explain how to select the appropriate margin of safety, the weighting factor, or any of the other variables,” the court wrote. “All that it provides is a formula for computing an updated alarm limit.”

And that was a problem because the Supreme Court had previously ruled that mathematical formulas couldn’t be patented. In the 1972 case of Gottschalk v. Benson, the high court rejected a patent on a method for converting numbers from conventional binary format to a binary-coded decimal format.

“Reasoning that an algorithm, or mathematical formula, is like a law of nature, Benson applied the established rule that a law of nature cannot be the subject of a patent,” the court wrote in the Flook decision.

Dale Flook argued that his patent was different because he had combined a computer program with a real-world industrial application. But the Supreme Court found this reasoning unpersuasive.

“The notion that post-solution activity, no matter how conventional and obvious in itself, can transform an unpatentable principle into a patentable process exalts form over substance,” the court wrote. “The Pythagorean theorem would not have been patentable, or partially patentable, because a patent application contained a final step indicating that the formula, when solved, could be usefully applied to existing surveying techniques.”

There’s an important ambiguity in the Benson and Flook decisions. They talk about algorithms and mathematical formulas, not about software or computer programs. If you treat “algorithm” and “computer program” as synonymous, then these two decisions amount to a flat ban on patenting software.

It’s not clear if this is what the Supreme Court had in mind. Later rulings from lower courts tried to draw a distinction between mathematical algorithms on the one hand and algorithms that manipulate more “concrete” types of data, like audio or video, on the other. Other rulings have suggested that software techniques for improving computer performance are patentable.

Yet in both the Benson and Flook cases, the Supreme Court suggested that it was up to Congress, not the courts, to extend patent protection to software. And Berkeley legal scholar Pam Samuelson points out that the Supreme Court’s Benson ruling cited a 1966 report by a presidential commission recommending that software not be eligible for patent protection.

In 1981, the Supreme Court decided a third case on software patentability: Diamond v. Diehr. The high court approved a patent on the idea of using a computer to manage a rubber-curing process, dynamically calculating the optimal time to open the press in order to produce perfectly cured rubber parts.

This ruling is often described as the Supreme Court overruling its prior decisions and opening the door to software patents. But it might be better understood as representing the other side of the Flook coin. Flook’s patent was rejected because it merely claimed a mathematical algorithm. The Diehr patent, by contrast, was allowed because it claimed a lot more than just an algorithm.

“The respondents here do not seek to patent a mathematical formula. Instead, they seek patent protection for a process of curing synthetic rubber,” the court wrote. “Their process admittedly employs a well-known mathematical equation, but they do not seek to pre-empt the use of that equation.”

In other words, rather than reading the Diehr decision as saying that software is patentable, it makes more sense to read it as allowing the patent because it wasn’t a software patent.

Still, the Diehr decision did help to crack open the door to more software patents. But it’s notable that this 1981 decision was not followed by an avalanche of software patents. Software patenting continued to be a relatively rare phenomenon throughout the 1980s. That would change in the 1990s, however.

Software patents expand

President Ronald Reagan signs the Federal Courts Improvement Act, which created the United States Court of Appeals for the Federal Circuit, on April 2, 1982.

Credit: Photo courtesy of the Federal Circuit Historical Society

Perhaps the most important development for the expansion of software patents came in 1982 when Congress created a new, specialized appeals court for patent cases, called the “US Court of Appeals for the Federal Circuit.” This was a departure from the system of generalist courts that handle almost all other areas of law in the United States.

Congress thought a specialist patent court would be better suited to handling the complexities of patent law. For 15 years after the creation of the Federal Circuit, the Supreme Court largely left the lower court free to shape patent law with minimal interference from above.

“There are no substantive [Supreme Court] patent decisions between 1982 and 1997,” Lemley said.

As a result, the Federal Circuit started to think of itself as the “Supreme Court of patents”—and it began to push back against the Supreme Court’s limits on patenting software.

At first, the Federal Circuit did this cautiously, finding ways to shoehorn software patents into the legal framework the Supreme Court had established in the 1970s. In a landmark 1994 case, the Federal Circuit approved a patent on the concept of using anti-aliasing software to smooth out the waveforms on a digital oscilloscope, reasoning that “a general-purpose computer in effect becomes a special-purpose computer once it is programmed to perform particular functions.”

When this and other rulings didn’t result in a rebuke from the Supreme Court, the Federal Circuit became bolder. In 1998, the court approved a patent on software-based strategies for managing mutual funds.

“Today, we hold that the transformation of data, representing discrete dollar amounts, by a machine through a series of mathematical calculations into a final share price, constitutes a practical application of a mathematical algorithm, formula, or calculation, because it produces a useful, concrete, and tangible result,” the court wrote.

This reasoning seems almost impossible to square with the reasoning that the Supreme Court used two decades earlier. After all, the “alarm limit” calculated by Dale Flook’s software invention was also a “useful, concrete, and tangible result.” In effect, the Federal Circuit overruled the Supreme Court—something that isn’t supposed to happen. But the Supreme Court didn’t review the decision, so it became binding precedent for the next decade.

But in the mid-2000s, the high court started taking a closer look at the Federal Circuit’s work—and justices didn’t like what they found.

The Supreme Court strikes back

Justice Anthony Kennedy wrote for the court in Bilski v. Kappos.

The Supreme Court fired its first shot across the Federal Circuit’s bow in 2006. It considered a patent on the concept of measuring the level of an amino acid called homocysteine in a patient’s blood as a way of determining whether the patient has a vitamin deficiency. The Supreme Court ultimately decided not to issue a formal ruling on the case. But the decision to hear the case at all signaled a renewed interest in the issue of what “inventions” could be patented—a topic the high court had ignored for decades. Lawyers started raising issues of patentable subject matter not only in biotechnology cases but in software cases as well.

In 2010, the high court reviewed a patent on the concept of hedging against the risk of price fluctuations in commodity markets. This patent was too much even for the Federal Circuit, which had rejected it as too abstract. In its Bilski v. Kappos ruling, the Supreme Court agreed, rejecting the patent in a unanimous vote.

This should have been a clear signal to the Federal Circuit that it should rethink its broader approach to patentable subject matter. But the Federal Circuit was slow to get the hint.

After handing down its Bilski ruling, the high court then asked the Federal Circuit to reconsider a pending case called Mayo v. Prometheus related to a medical diagnostic test that effectively claimed a fact about the human body. The Federal Circuit had previously found the patent valid, and when it looked at the case a second time… it again found the patent to be valid. The Supreme Court wasn’t amused. The high court took the case and overruled the lower court in a unanimous decision, finding that the patent claimed an unpatentable law of nature.

Remarkably, the Federal Circuit still didn’t get the hint. After its Mayo ruling, the high court sent another case, this one dealing with the patentability of human genes, back to the Federal Circuit for reconsideration in light of the new ruling. Once again the Federal Circuit stood its ground, finding for a second time that the gene patent was valid. Again, the Supreme Court took the case and overruled the lower court in a 9-0 ruling.

After these back-to-back smackdowns, some Federal Circuit judges started to get the message. But the court remained deeply divided. In 2013, the appeals court issued a fractured ruling in a software patent case called CLS Bank v. Alice. Five judges wanted to allow the patent. Five wanted to reject it. The ten judges participating in the case wrote a total of seven opinions.

But while the Federal Circuit thought the case was complicated, the Supreme Court didn’t. The justices took the case and ruled unanimously that the patent—which claimed a software-based method of making sure two parties to a financial transaction keep their promises by having a third party hold their funds—was too abstract.

“The relevant question is whether the claims here do more than simply instruct the practitioner to implement the abstract idea of intermediated settlement on a generic computer,” the court wrote. “They do not.”

The software patent smackdown had arrived in force.

Rejection

While it might seem obvious that such generic “do it on a computer” patents shouldn’t be valid, the reality is that there are thousands of other patents like it. A decade earlier, patent defendants would rarely bring patentability questions because the Federal Circuit had been rejecting them so decisively. But after the Alice ruling, defendants brought these arguments up in almost every software patent case—and they kept winning.

Last year, the legal scholars Paul Gugliuzza and Mark Lemley surveyed all of the cases on patentable subject matter that came to the Federal Circuit in the first three years after the Alice ruling. There were 104 of them—98 focused on software patents and six others focused on biotechnology. Trial courts rejected patents in a remarkable 96 percent of these cases. The Federal Circuit was more forgiving but only slightly; it rejected patents 92 percent of the time.

Pam Samuelson says that the Alice ruling has reduced frivolous patent litigation.

“A lot of patents that issued during the ‘everything’s patentable’ era during the 1990s and 2000s are just not being asserted in the manner that they were before,” she told Ars. “If they’re being asserted, the expectations of being able to extract large licensing fees have diminished.”

While the Federal Circuit has rejected a lot of obviously bogus patents, it has also made it clear that there is still significant room for patents on software. A good example is the 2016 ruling on a case in which a small software company called Enfish sued Microsoft for infringing a patent that claims a method for organizing data in a database.

Fundamentally, the Enfish patent claimed a way of organizing data in computer memory. That seems a lot like the very first software patent the Supreme Court rejected in 1972, which claimed a method for converting between number formats in memory.

But the Federal Circuit argued that these are different because the Enfish patent is “directed to a specific improvement to computer functionality,” whereas the patent in the 1972 Benson case claimed “an abstract mathematical formula on any general-purpose computer.” But the line between these definitions isn’t clear. Benson’s method for converting between binary number formats surely made computers operate more efficiently, for instance, and the Enfish patent claims abstract data structures that can be implemented on any general-purpose computer.

After repeated smackdowns, the Federal Circuit has clearly gotten the message that the Supreme Court would like fewer software patents. At the same time, the lower court is up to its old tricks, looking for ways to allow as many software patents as possible within the confines of Supreme Court precedent.

Thus, if the Supreme Court goes back to ignoring patent law—and there haven’t been any significant rulings on subject matter eligibility since Alice four years ago—we’re likely to see the same gradual expansion of software patenting that we saw in the 1980s and 1990s.

The underlying argument

Beneath the perennial institutional tension between the Supreme Court and the Federal Circuit lies a philosophical disagreement. Federal Circuit judges spend a lot of time around patent lawyers, and patent lawyers are hostile to the very idea of cordoning off some types of discoveries from the patent system. Supreme Court justices, on the other hand, are not patent specialists and tend to be more cognizant of the downsides of excessive patent protection.