To many, a computer program is a mystifying series of (seemingly) random numbers, letters, and symbols that come together to produce extraordinary results that improve our lives every day.

Those who write computer programs are often viewed by the less technologically-savvy as gifted authors and designers whose efforts deserve protection under the law. A question less frequently contemplated is what kind of protection is that. What if the program simply carries out a computer function, or itself creates something new? Are either of these programs copyrightable?

Congress protected computer programs under copyright on the theory that computer programs are made up of words and numbers and therefore can be classified as “literary works” under the definition in §101 of the Copyright Act. The difficulty, of course, is that computer programs also - and primarily - perform functions in computers.

Seeking to protect computer software through copyright raises a fundamental tension: computer programs are essentially machines. The Copyright Act specifically says that copyright protection does not extend to any idea, process, system or method of operation. 17 U.S.C. §102(b). The classic protection for functional inventions lies under the Patent Act, not the Copyright Act.

The history of the protection of computer programs is a history of courts trying to resolve this tension in the Copyright Act. A new chapter in that history book was recently written in Automated Management Systems, Inc. v. Rappaport Hertz Cherson Rosenthal, P.C.(S.D.N.Y.), where a copyright infringement claim directed to protecting the designs of a computer program and an associated database survived a motion to dismiss that challenged their copyrightability.

In this article, we discuss the progression of computer program protection, culminating in the Automated Management decision. If followed, this decision blurs the lines of copyrightability of computer database compilations, and we consider what that will mean for artificial intelligence and future computer programs.

Background

The issue of classifying protection arose early on in the computer program era, exemplified by a case called Whelan Assoc., Inc. v. Jaslow Dental Lab., Inc. decided in 1986 by the federal appeals court in Philadelphia. Copyright protection does not extend to an “idea,” but only to an author’s expression of an idea. There is, however, some difficulty in drawing the line between idea and expression: How specific does the development of an idea have to be before it attains the level of protectable “expression”?

In Whelan, the Third Circuit drew the line between idea and expression at a very high level of generality. In substance, the court held that the idea of a software program automating the operations of a dental laboratory was an unprotectable idea, but the structure, sequence, and function of any computer program implementing that idea could be copyrightable expression. This decision, however, effectively protected functional elements, contrary to the process-system-method of operation exclusion of §102(b).

A few years later, the federal appeals court sitting in New York took a different and more restrictive approach to software protection. In Computer Associates Int’l, Inc. v. Altai, Inc.,982 F.2d 693 (2d. 1992), the Second Circuit held that a court had to distinguish and put aside the functional and therefore unprotectable elements of software from those that are protectable under copyright before determining infringement. Altaiheld that the courts should apply what it called the abstraction-filtration-comparison test to computer programs to determine infringement. This test requires the court to identify the potentially protectable features of the program and for the judge to determine whether they are unprotectable because they are dictated by efficiency or by external factors – features that a program of that kind must have – or material in the public domain. Only then, after identifying the protectable elements, does the court determine infringement.

Automated Management’sCompilation Protection

Despite Altai’sreduction of copyright protection of the structure, sequence, and function of a computer program, protection has been extended to software products under the heading of “compilation,” including now in the Automated Managementcase.

In copyright law, a compilation is an original selection, coordination, or arrangement of materials. The underlying materials themselves need not be copyrighted. 17 U.S.C. §101. The copyrightable authorship in a compilation lies in the originality of the selection and arrangement of the items compiled. Thus, for example, the array of common baseball statistics an author selected to predict the outcome of a game has been held to be copyrightable authorship.

In the Automated Managementcase, the software involved was a program called “Landlord and Tenant Legal System.” As its name suggests, the software was used to automate the operations of a landlord-tenant law firm. The law firm had experienced poor performance from Automated Management’s program and wished to replace it with a new system having the same functionality.

The manufacturer of the new software created a program modeled on Automated Management’s program’s functionality, and was sued by Automated Management for copyright infringement. Also among the defendants was the law firm, accused of violating its end-user agreement by giving access to the software to another software manufacturer.

The Automated Management program had three components alleged to have been infringed: an email alert function, a data entry function, and a “data dictionary.” The first two features are self-explanatory.

The data dictionary was the database-related part of the program. It is described in the case as explaining the structure of the software database. The database dictionary lists the elements in the database, describes its data type and defines what it is. Automated Management conceded that the “database schemas” in the defendants’ database were different from plaintiff’s, but claimed that the defendants’ database structure copied substantial portions of plaintiff’s.

The court noted that the data dictionary arranged 2,213 types of data into 51 tables, an arrangement claimed to be original to Automated Management. The court said that while the idea of arranging data into a database cannot be copyrighted, a unique design arrangement can be a “compilation.”

The defendant argued that its own software did not utilize a data dictionary as such and therefore could not have infringed by copying. The court responded that the defendant copied the database structure. The court pointed to “the appearance in both databases of tables containing the same sets of variables, identified by nearly identical names, and used for the same purpose.” Thus, despite the absence of a data dictionary in defendant’s software, it was reasonable to allow a jury to consider whether the plaintiff’s compilation of database elements was both original and infringed by defendant’s database using substantially the same elements.

Noteworthy is that this case concerns the database design, not the data itself; the data in the database was law firm data about its cases. The case therefore does not speak to the copyrightability of data itself. What the case does stand for is that a database design can be copyrightable and can be infringed by copying the selection of elements in the database and their arrangement.

Looking Forward

What this case teaches is that the line between the selection and arrangement of elements of a database and an uncopyrightable system or process is not easy to draw. Automated Management, if followed, is important because of the importance of databases in the world of artificial intelligence.

Artificial intelligence systems must train on data, must be tested on data and, to be useful, must be applied to data. How that data is selected and arranged and organized in a database may be protectable and may therefore be an element to be considered in analysis of intellectual property questions relating to artificial intelligence.

Further, what the Automated Managementcase illustrates – assuming that it is good law – is that protection of functional software can be achieved to some degree beyond the level of the code making up the program if the elements included in it and their arrangement is original. What is probably still not protectable is the functionality of a computer program – how the steps are performed in what order and what results the program provides. However, there is language in Automated Management that suggests that the selection and arrangement of functions could be a protected compilation, an approach that would come very close to protecting a system or process contrary to the Copyright Act.