“What’s past is prologue.”—Shakespeare, “The Tempest”
“History is more or less bunk.”—Henry Ford
Like the disputants in the old Certs commercial, they’re both right. The law of intellectual property relating to artificial intelligence and its products is both already established in the existing law, and will have to be invented as AI plays our games, writes our books, plays, and music and, with expert software, puts lawyers out of business.
Perhaps its complexities will be so difficult that we will have to resort to AI judges.
However, another sage tells us, “It’s hard to make predictions, especially about the future.” It is useful to look at what the law, particularly copyright law, already says about ownership and use of AI rather than trying to forecast what the law will say about the works of autonomous automatons.
Before commenting on copyright protection, note that anything that gives a business an advantage over competitors can be protected as a trade secret. This includes all aspects of AI.
The catch is that the information must be kept secret and is not protected against discovery or use by others if they learn the information without discovering it unfairly by breaching the secrecy. Reverse engineering, for example, is permitted. Getting the secret from an unfaithful employee is not permitted. Secrecy can be maintained while allowing users to use a device or service that contains the secret either by agreement or by security measures.
The U.S. Court of Appeals for the Eighth Circuit has held that an algorithm—for making a diagnosis or determination based on research, for example—embodied in documents can be copyrightable. However, it is doubtful that the underlying method, as separate from the specific words of the documents embodying the method, is copyrightable.
Computer programs are protected by copyright. The scope of protection is, however, relatively narrow, limited to little or nothing beyond the instructions themselves.
The structure, sequence, and organization of a computer program are generally considered to be uncopyrightable because systems or processes are not protectable by copyright (although the structure, sequence and organization of a package including programming can be copyrightable—see below).
The April U.S. Supreme Court decision in Google LLC v. Oracle America Inc. concerned copyright in application programming interfaces (APIs). The Federal Circuit decided that the “overall structure of Oracle’s API packages is creative, original, and resembles a taxonomy,” but the Supreme Court only ruled on whether Google’s use of the APIs was fair use and therefore not copyright infringement. The Supreme Court did find fair use.
Data and Databases
Data that is created—not merely collected—can be copyrightable if creative judgment goes into each datum. Merely collected data is not copyrightable unless the criteria for selecting the data to collect or the method of arranging it once collected constitutes creative authorship.
Estimates of used car values or coin prices have been held copyrightable. The Ninth Circuit has found copyrightable authorship in a list of names and addresses chosen, using judgment, from “multiple and sometimes conflicting sources.”
Databases, when copyrightable, are generally protected by copyright law as “compilations.” Under the Copyright Act, a compilation is defined as a “collection and assembling of preexisting materials or of data that are selected in such a way that the resulting work as a whole constitutes an original work of authorship.”
Copyright protection for selection and arrangement of data is very thin, however, prohibiting only “’bodily appropriation of expression,’ or ‘unauthorized use of substantially the entire item,’” the Ninth Circuit said. Even a match rate of 80% of the data has been held insufficient.
Data structures are copyrightable if there is creativity in the data structure and its organization. If the ideas behind the structure cannot be used without the structure implementing them, however, or if the structures are dictated by industry factors, the structure will not be copyrightable.
Protection, as with databases, is thin; virtually literal copying of the structure is required.
This is a highly unsettled area and one which is likely to give rise to great controversy as AI does more and more creative work. The contest currently is essentially between the AI machine owner and the user.
Although there is no decision clearly awarding copyright to a program owner in output resulting from use of the program by another, there are decisions that entertain that possibility if the program does the “lion’s share” of the creative work.
Systems applying AI may qualify for patent protection. The U.S. Patent and Trademark Office, Class 706 of patentable inventions is “Data Processing: Artificial Intelligence,” described by the Patent Office as “a generic class for artificial intelligence type computers and digital data processing systems and corresponding data processing methods and products for emulation of intelligence.”
The law already provides for protection of many of the components of AI, even allowing the patenting of AI systems themselves. It is reasonable to hope that the law will meet the challenge of artificial intelligence in ways that we cannot now anticipate, for, as Justice Oliver Wendell Holmes said, “The life of the law has not been logic; it has been experience.”
This column does not necessarily reflect the opinion of The Bureau of National Affairs, Inc. or its owners.
David Rabinowitz is partner at Moses & Singer and a member of its Litigation, Intellectual Property and Law Firm Management practice groups.