Updated July 13, 2020: 

Algorithm patents are patents for math that describe a particular process and purpose. This means that the patented algorithm can't just cover all particular uses of an algorithm, but rather must cover a single particular algorithm. One common test is to turn the algorithm into a single electronic circuit, which must differentiate itself from other circuits. Since these kinds of algorithms make our computerized society function, these patents are a necessary part of maintaining that society.

At the same time, the various mechanics of protecting intellectual property have led to an extensive debate on all levels of society, from the local to the global. Examples like the European Union's attempted draft of a Directive on the Patentability of Computer-Implemented Inventions demonstrates how views among member nations can vary wildly. This is complicated by the very nature of information technology, as the internet makes processes globally accessible while patent protections remain rooted in individual nations.

Over time, algorithm patents are becoming more and more common, with applications increasing eight times in four years from 2011 to 2015. With no sign of slowing, it is likely that algorithm applications will become one of the biggest patent categories in the next few years. This in turn will probably continue creating a backlog of patent applications in the field, making it more difficult for late entrants every day.

Patent Code Classifications

Most of the top computer companies are now actively involved in algorithm patenting. IBM is a leader in raw patent numbers, but Samsung and Qualcomm are serious contenders as well. Chinese companies are establishing a strong presence in the patent field, with many companies and schools in the top 40 applicants.

As far as what kinds of things the algorithms cover, top patent code classifications (CPC codes) include:

  • Learning machines, also known as artificial intelligence
  • Edge detection and segmentation
  • Medical diagnosis device software
  • Image inspectors for biomedical devices
  • Networking components

Many of these items relate to the emergence of the so-called Internet of Things, where interconnected smart devices increasingly dominate previously stable appliance markets. Clearly there is a race between domestic companies and Chinese firms to control the underpinnings of the software for this radical new concept.

At this point in history, most countries recognize computer programs as copyrightable objects. This simplifies international copyright enforcement, as everyone is playing from the same handbook. Since copyright generally extends protection much longer than a patent, the use of copyrights preserves the inventor's control in ways that patents cannot. But patents also offer certain protections absent in an expression-based copyright, such as the defense of methods and ideas, and therefore remain in use.

Without the united body of law attendant to copyright, patents have to be applied for on a country by country basis. Legal requirements vary greatly but usually include a fee of some sort, substantial information sent to the patent authority, and public disclosure of the invention. The complexities of international patent law almost always require the retaining of a lawyer, making it an expensive process. Though it may seem like a lot of trouble, the patent will protect the ideas behind the algorithm, whereas a copyright only protects the algorithm itself.

A patentable invention meets five significant criteria:

  • The patented invention must be constructed of patentable items and subjects.
  • The inventions have to be usable in an industrial or another useful context.
  • The invention must be original.
  • The invention has to be inventive, or not obvious.
  • The patent paperwork must meet the requirements of the patent office.

In terms of software and algorithms, the requirements to be non-obvious and a patentable subject matter are the most important. Since the definition of invention varies from nation to nation, whether an algorithm is patentable at all is often a source of contention. Emphasizing the technical aspects of the invention will help here. As far as the non-obviousness requirement, the idea is usually to produce inventions that contribute something new.

Of course, the true test of any legal instrument is the courtroom, and current cases over algorithm patents and copyrights will end up shaping the protections environment going forward. As these cases resolve, new legal requirements will inevitably emerge.

If you need help with algorithm patents, you can post your legal need on UpCounsel's marketplace. UpCounsel accepts only the top 5 percent of lawyers to its site. Lawyers on UpCounsel come from law schools such as Harvard Law and Yale Law and average 14 years of legal experience, including work with or on behalf of companies like Google, Menlo Ventures, and Airbnb.