How to Make Sure the Owner of Your Intellectual Property Is You and Not Your AI

AI systems can simplify many complicated tasks. They can also complicate simple tasks, such as identifying who is the author, for intellectual property ownership purposes, of a new creation. AI systems can fill out forms, recommend songs, and complete hundreds of other tasks, but even the most sophisticated AI can’t fill in the blanks when asked to identify the names of the inventors or authors, even if AI was involved in the creation of an invention or work. Although identifying the most effective types of intellectual property law to protect inventions, designs, or expressions is not a new challenge, the introduction of AI in the creative process requires some careful thinking about choosing an intellectual property protection strategy.

When patent law and copyright law came into being, it was very clear that inventors and authors were natural persons, because there wasn’t anything else that was up to the task. The world is different now, in that AI can do some things that might be considered inventive enough to patent and can create works that might be considered creative enough to be copyrightable. In those cases, who (or what) gets listed on the patent application or copyright registration can get complicated. Patent and copyright statutes generally don’t recognize non-human inventors and authors.

However, just because an AI system created something without ongoing human intervention, that doesn’t mean the result isn’t novel, expressive, or otherwise worth protecting with exclusive rights. And it can take considerable effort and investment to get an AI system to the point where it can create inventions and works. Companies investing in AI efforts would do well to plan ahead to determine what intellectual property rights can be used to protect the fruits of those investments.

There is no easy solution or one-size-fits-all approach. Different types of intellectual property rights can overlap, but often one type of intellectual property right is sufficient to exclude competitors from using whatever is protected by that right. The best fit may be a combination of intellectual property rights, in order to capture different aspects of the commercial applications of the creation and to accommodate changes over the lifecycle of those commercial applications.

A person in 1908 who wanted to produce a board game “provided with corner-spaces, intervening spaces of different denominations, and distinguished by distinctive marking, so that the corresponding divisions on the different spaces may be recognized, movable pieces having reference to the spaces, a chance device to control the movement of the pieces, checkers, and tickets representing money, deeds, notes, mortgages, bank mortgages, charters, legacies, and luxuries” would be excluded from selling it without the permission of the owner of U.S. Patent 748,626. Eventually, that patent protection expired, but the well-known Monopoly™ board game is still heavily protected—now by copyright and trademark law covering the wording, visual design, and branding.

A strategy of patching together protections for different phases of the life cycle, or for different aspects of a creation and its commercial applications, can be used today with AI systems. The usual challenges of fitting within the various intellectual property regimes still apply—if the key asset can’t be commercialized without revealing the inner workings, trade secret protection alone won’t work; if the work is the result of plenty of effort but little creative expression, copyright alone won’t work; if the lifecycle of the target product will take decades to get to market, patent law alone won’t work. And if outside vendors or multiple employees contribute to development, training, or creative output, companies need to be sure that contract and employment terms support their preferred intellectual property protections.

AI introduces a new constraint to work around—the questions of inventorship and authorship. If the output of an AI program is something that would be clearly patentable if created by mental efforts of a natural person but not if done by computational means, all is not lost. Certain aspects of the creation and its commercial applications may still be patentable. And copyright, trade secret, or trademark protection can be used to help fill in the gaps. Companies that have organized their investments and legal strategies around the creation of patentable inventions will need to start thinking, sooner rather than later, about what steps are needed to secure other types of intellectual property protection as well.