If you write a book or a song, you can get copyright protection for it. If you invent a pill or a better mousetrap, you can patent it. But for decades, software has had the distinction of being eligible for both copyright and patent protection.
The patent system, critics say, is complex and expensive — and a bad fit for the fast-moving software industry. And they wonder, doesn’t the protection offered by copyright suffice?
On Monday, for the first time since 1981, the Supreme Court will hear arguments on whether software — or more precisely, computer-implemented inventions — can coninue to be patented. The case, Alice Corp. v. CLS Bank, focuses on software built to eliminate “settlement risk” in currency and financial transactions where money is held in escrow and one party could renege on a deal and leave the other holding the bag. Alice Corp. claims that CLS Bank offers a service that infringes on its patent.
The first time the high court took up such a case, it ruled that “one may not patent an idea” or something as abstract as an algorithm.
Alice Corp. has argued that because its innovation requires the use of a computer, it amounts to an invention that is patentable.
In the 1970s, the high court placed strict rules on software-related patents. Since then, a lower court has effectively overruled its precedents, flooding the marketplace with software patents.
While the arguments in this debate have barely changed since the 1970s, the players have changed radically — in point of view, at least. In 1972, IBM was a leading software patent opponent. Today Big Blue is one of the concept’s biggest supporters. In 1991, Bill Gates warned that patents could bring the software industry to a “standstill.” Today, Microsoft is fighting to protect the tens of thousands of software patents in its portfolio.
Given the business interests at stake, the high court faces a rich dilemma: Since 1998, the U.S. Patent and Trademark Office has issued hundreds of thousands of software patents — collectively worth billions of dollars to companies such as IBM, Microsoft and Oracle. If they rule that software cannot be patented, the justices could decimate the patent arsenals of some of America’s wealthiest companies.
The first high-profile battle over software patents pitted two of America’s most powerful companies, AT&T and IBM, against each other. Gary Benson and Arthur Tabbot were researchers at AT&T’s legendary Bell Laboratories, and in 1963 they sought a patent on a method of converting between two different binary formats.
Their invention worked like this: Suppose an AT&T customer dials “202.” In binary, 2 can be represented as “0010” and 0 as “0000.” So as the customer dials “202,” AT&T’s computer system might store this as “0010 0000 0010.” This format is known as “binary coded decimal.”
But it’s more efficient for a computer to represent 202 in its “pure binary” form: “11001010.” Benson and Tabbot had developed a sequence of mathematical operations that allowed computers to rapidly convert a binary-coded decimal number such as “0010 0000 0010” into the more compact form “11001010.”
The Patent Office rejected their application, arguing that one could not patent “a logical list of mental steps.” Mathematical formulas had long been off-limits for patent protection. Benson and Tabbot’s application was not a patentable invention because its sequence of mathematical operations was simply carried out by an electric circuit rather than with a pencil and paper.
When the case reached the Supreme Court in 1972, IBM — then the world’s largest computer company — filed a brief opposing the patent and software patents in general.
Until 1968, IBM had given its software away for free to customers who bought its hardware. In the company’s view, allowing patents on software would “have the inevitable effect” of “stifling developments in computer programming” — thereby limiting the market for IBM’s expensive mainframe computers.
IBM also argued that copyright protection was a better fit for the software industry than patents. “The need for protection against copying is very real,” IBM conceded.
But crucially, copyright protection allows someone to independently develop software to achieve the “same overall result” as a copyrighted program. In contrast, patent law does not, thereby limiting the opportunities of future innovators and creating the risk of accidental infringement and wasteful litigation.
While IBM made the case against software patents, a fierce adversary was making the case for them: a trade group called the Association of Data Processing Service Organizations, which represented some of the first software companies.
Its members were looking for ways to break Big Blue’s stranglehold over the software business. They were enthusiastic supporters of the antitrust lawsuit the government filed against IBM in 1969.
Marty Goetz was an executive at the software company ADR, which produced Autoflow, one of the first commercial software products.
“For five years, we were selling our software and IBM was giving away their software,” Goetz says. “Autoflow ended up having a competitor: a very inferior product by IBM.” Goetz says IBM’s pricing policy and promises of future improvements made customers reluctant to buy Autoflow. “That was one reason we got a patent,” Goetz said.
At Goetz’s urging, the association filed a brief in favor of software patents. Its arguments are still being repeated by software patent advocates today.
It worried that if its members produced a successful software product, “IBM and the other hardware manufacturers will be able to copy the technological concepts of the priced software and put out their own version” for free — a “devastating prospect.”
The association also objected to the government’s (and IBM’s) argument that software patents were patents on mental steps. When you load a computer program into a computer, they said, you effectively turn it into a new type of machine eligible for patent protection.
The Supreme Court sided with the government and IBM. “One may not patent an idea,” Justice William Douglas wrote for the court. So Benson and Tabbot’s patent “in practical effect would be a patent on the algorithm itself.”
The high court doubled down on this stance six years later. An inventor named Dale Flook applied for a patent covering a new formula for setting an “alarm limit” during a catalytic conversion process. An alarm limit signals that a variable such as temperature or pressure has reached an unsafe level.
The Supreme Court had rejected the Benson patent because it claimed all conceivable applications of the BCD-to-binary conversion formula. Flook argued that his own patent was different for two reasons. First it was limited to a specific industrial application: the catalytic conversion of hydrocarbons. And second, unlike in the Benson case, the computer in the Flook process does something tangible with the result of the calculation: modifying the limit that triggers safety alarms.
Flook was effectively seeking ownership of a mathematical algorithm. The Supreme Court said no.
In 1981, the Supreme Court finally found a computer-related patent it liked.
To make rubber products, raw rubber is placed in a molding press and heated. Success depends on opening the press at precisely the right time. Inventors Thomas Diehr and Theodore Lutton realized that they could improve the rubber-curing process if they constantly measured the temperature inside the press, fed the data into a computer, and used it to recalculate the timing.
The patent office wanted to reject the patent based on the Supreme Court’s previous rulings. But a narrow majority of five Supreme Court justices disagreed. Justice William H. Rehnquist wrote that Diehr and Lutton “do not seek to patent a mathematical formula. Instead, they seek patent protection for a process of curing synthetic rubber” that happens to employ a mathematical algorithm as one of its steps.
The decision complicated the job of the patent office. It did not overrule the Flook and Benson decisions; the kind of “pure” software patent rejected in those was still illegal. Inventions that tied a computer program to a real-world application, like Diehr and Lutton’s rubber-curing technique, became permissible. But no one was sure where to draw the line between the two.
“We had this anarchy about how these decisions would be interpreted in the office,” says Bruce Lehman, who led the patent office starting in 1993. In the decade after the Diehr decision, courts offered little to clarify which computer-related inventions were patent -eligible and which were not.
As the patent office allowed more software patents, many in the industry reacted with alarm. Companies worried that patents would spark unnecessary litigation.
“If people had understood how patents would be granted when most of today’s ideas were invented, and had taken out patents, the industry would be at a complete standstill today,” Microsoft chief executive Bill Gates wrote in a 1991 memo to his executives. Microsoft was still a relatively small company in 1991, and Gates worried that “some large company will patent some obvious thing,” which could give it “a 17-year right to take as much of our profits as they want.”
Gates took a pragmatic approach; the company got to work filing hundreds, and eventually thousands, of patent applications. Other software companies began to cry foul.
In 1994, Lehman organized public hearings on software patents and invited leading software companies to comment. The database company Oracle argued that patent protection was “not appropriate for industries such as software development in which innovations occur rapidly, can be made without a substantial capital investment and tend to be creative combinations of previously-known techniques.” Oracle’s views were echoed by Adobe, creator of Photoshop.
IBM, on the other hand, had changed its mind. “As the industry matures and competition from overseas increases, patents will be the key to protecting the most valuable U.S.-originated innovations,” Big Blue argued.
Microsoft also endorsed software patents, despite the reservations of its chief executive.
In 1982, Congress made a seemingly innocuous change to the structure of the court system that had a profound impact on the legal status of software patents.
Concerned that patent law had become too complex for generalist judges, Congress created the U.S. Court of Appeals for the Federal Circuit and gave it jurisdiction over all patent appeals. Perhaps because its judges spend so much time rubbing elbows with patent attorneys, the new court would prove to have a strong pro-patent bias.
An important turning point came in 1998. State Street Bank fought for a patent on the use of a computer to implement a strategy for managing mutual funds. The Federal Circuit approved it.
“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 ruled.
The State Street decision seemed to ignore the Supreme Court’s views altogether. Pam Samuelson of the University of California at Berkeley’s law school, says it’s “not possible” to square the State Street ruling with the Supreme Court’s precedents. In her view: “They didn’t like the ruling, and so they gave it a narrow interpretation. In effect, they overruled it.”
Unsurprisingly, the number of software patents soared in the 1990s — as did the number of lawsuits involving software patents.
At first, the Supreme Court let the Federal Circuit shape patent law with minimal interference. But by the time John G. Roberts Jr. became chief justice in 2005, the Federal Circuit’s patent-friendly jurisprudence had produced record levels of patent litigation and an avalanche of bad press. The Supreme Court gave the Federal Circuit closer scrutiny.
The high court didn’t like what it found. The first few patent rulings of the Roberts era were unanimous or near-unanimous reversals of the lower court’s decisions.
In a 2009 oral argument, Roberts commented that lower courts “don’t have a choice” about following Supreme Court precedents. “They can’t say, ‘I don’t like the Supreme Court rule so I’m not going to apply it.’ ” Then, he added wryly: “other than the Federal Circuit.”
In 2012 the high court unanimously overruled a Federal Circuit decision allowing patents on medical diagnostic techniques. Last year, the high court rejected patents on human genes, which the Federal Circuit had previously approved.
On Monday, the Supreme Court takes up Alice Corp v. CLS Bank.
Like the mutual fund patent that the Federal Circuit upheld in 1998, the process described by Alice’s patent takes place entirely inside the computer and isn’t tied to any specific machine or industrial process.
Alice argued that the fact that the patent required the use of a computer was sufficient to make it a patentable invention. When the case reached the Federal Circuit, the court agreed in a 2 to 1 decision.
The dissenting judge, Sharon Prost, was furious. “The majority has failed to follow the Supreme Court’s instructions — not just in its holding, but more importantly in its approach,” she wrote. “Just a few months ago, the Supreme Court reversed us for a second time in its last three terms, hinting (not so tacitly) that our subject matter patentability test is not sufficiently exacting.”
Her dissent prompted consideration by a larger panel of Federal Circuit judges. The result was a mess, as no single opinion gained majority support. Five judges wanted to invalidate the patent outright. The other five judges wanted to uphold at least some of the claims.
In a dissenting opinion signed by three colleagues, Judge Kimberly Moore worried that rules against patenting mathematical algorithms and other abstract ideas would be “the death of hundreds of thousands of patents, including all business method, financial system and software patents as well as many computer implemented and telecommunications patents.”
Critics say that’s the point: The Supreme Court placed limits on software patents in the 1970s, and the Federal Circuit is supposed to enforce them. In their view, business method and software patents shouldn’t have been granted in the first place.
And while a decision invalidating thousands of software patents could be bad for the bottom lines of some large software companies, it could be good for the industry as a whole. Litigation by “patent trolls” cost defendants at least $29 billion in 2011, one study showed. Those costs are driven by software patents.
The arguments about software patents in the 1970s and 1990s were theoretical. Opponents predicted that software patents would have a detrimental impact on software innovation. But the concept was so new that they couldn’t know for certain.
Now the evidence is in. Today IBM, Oracle, and Adobe all have thousands of software patents, and unsurprisingly they’re no longer opposed to software patents. But software patents have had all the negative effects they anticipated: Innovative companies have been forced to divert resources from hiring engineers to hiring patent lawyers. Large companies no longer on the cutting edge have been using broad patents to demand cash from small, more innovative companies. Many programmers and software entrepreneurs view patents as more a nuisance than a reward for innovation.
The Supreme Court could well settle the matter this spring. The question is, will it draw more inspiration from IBM’s 2014 brief making the case for software patents — or from IBM’s 1972 brief opposing them?