By Lauren J.F. Barta and Timothy E. Grimsrud
Artificial Intelligence is already used in healthcare in a multitude of ways, including robotic-assisted surgery, virtual nursing assistants, diagnostics, drug discovery, hospital workflow, consumer integrated health, and personalized medicine, and its use is only expected to grow.[1] AI technologies are able to sense, comprehend, act, and learn—this is different from typical systems that merely identify a condition and have a preprogrammed response to the condition. Indeed, AI technologies are anticipated to continue to evolve from automated intelligence, which automates existing steps in processes and may require human involvement, to autonomous intelligence, which is able to adapt and act without human involvement. As discussed below, the current U.S. intellectual property system is not equipped to address the implications of this evolution to AI as an autonomous actor.
AI as Inventor and Author
The U.S. Patent and Copyright Act do not currently account for inventions and works that are created entirely by AI.
The U.S. Patent Act presumes that only humans can be inventors. For example, it states that “[w]hoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor. . .,” defines “inventor” as being an “individual,” and explains that an “individual” must be a “human.” Based on this, the law does not allow AI to be an inventor, and therefore, AI may not obtain a patent. But, what about humans who use AI to create an invention? Can the human obtain a patent on the invention?
One view is that it depends on whether AI generated the invention on its own or whether a human conceived of the idea for the invention with the assistance of AI. If the AI generated the idea on its own, arguably the human would not be entitled to a patent on the idea. Prior to the America Invents Act (“AIA”), the Patent Act expressly stated that a “person” is not entitled to a patent if he or she did not invent the subject matter sought to be patented. Even after the AIA, there are proceedings meant to address the circumstance when a person attempts to obtain a patent on an invention he or she did not invent. However, another view is that even if AI generates an idea on its own, the human is truly the entity that conceived of the invention and is therefore the inventor. The AI is merely a tool that the human used to arrive at the invention.
One public example of this occurred in the 1990s, when scientist Stephen Thaler employed his AI system called the “Creativity Machine” to generate ideas. The Creativity Machine generated an idea on which Thaler ultimately succeeded in obtaining a patent. Moreover, while Thaler acknowledged that the Creativity Machine generated the idea, Thaler was the only named inventor on the patent. While it appears that the U.S. Patent Office was not aware of the Creativity Machine’s involvement in the invention, it is unclear whether or to what extent such involvement would have mattered to the Patent Office.
There are also public examples of AI generating what would otherwise qualify as copyrightable works. For example, an original painting created by a human is almost certainly considered a copyrightable work. Is the same true for art created by AI? The Christie’s auction house first sold an AI-created artwork in October 2018 titled “Portrait of Edmond Belamy” for $432,500. Can that work be copyrighted? The U.S. Copyright Office expressly stated that “[t]he office will not register works produced by nature, animals, or plants. . . Similarly, the Office will not register works produced by a machineor mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author. . .”
How will this standard apply to original artwork created by AI? While there is little law on this issue, the law that does exist does not look promising for AI. For example, on April 23, 2018, the Ninth Circuit Court of Appeals considered whether a photograph taken by a crested macaque could be copyrighted. The macaque took photographs of himself when a wildlife photographer left his camera unattended in the reserve in which the macaque lived. The photographer published the “monkey selfies” in a book and claimed the copyright over the photographs. PETA on behalf of the macaque brought a lawsuit against the photographer claiming copyright infringement. Ultimately, the Ninth Circuit held that the macaque did not have the ability to bring a suit under the Copyright Act because the Act does not authorize animals to file copyright infringement suits. Courts may take a similar view to works by AI.
In short, as the law currently stands, only humans may qualify as “inventors” and “authors” under the Patent and Copyright Acts. This restriction may present challenges in protecting potentially valuable AI-created intellectual property.
Ownership of AI Created Works
The original inventor or author is deemed to own a patented or copyrighted work unless the original inventor or author assigns his or her rights to another. Accordingly, because AI cannot be an inventor or author, AI cannot be an owner of patented or copyrighted work.
But the analysis does not end there: The likely upcoming legal fights will relate not to whether the AI itself is the owner of AI-created inventions and works but whether the AI developer or the company who integrates the AI into its system is the owner of the AI-created inventions and works. Not many companies have the capability to create AI, but many companies want to use AI in their own systems. When AI is integrated into a company’s system, depending on the abilities and limitations of the AI, the AI is likely to be trained using the information it obtains from the system. In addition, the company may provide feedback to the AI based on its output, which further trains and develops the AI and how it acts. Thus, when AI innovates, is the resulting work product that of the original AI developer or the user who trained the AI?
More and more AI developers are describing their AI as “transparent,” such that one can understand how the AI is operating and being trained by the system in which it is integrated. This may allow companies to trace why and how the AI acts and, therefore, whether the AI’s innovation is the result of the AI developer or its training.
Setting aside transparency issues, this is likely a fight that companies can resolve at the outset of their relationship. AI developers and healthcare and medical companies that may want to use AI in their systems should consider incorporating provisions in their agreements that outline ownership rights over AI-created inventions and works.
Infringement by AI
Under patent and copyright law, systems and machines have not typically been viewed as being “actors” capable of infringement. Instead, the human or company that was using the system and/or machine to create a patented technology or to reproduce a copyrighted work was deemed to be the infringer. But as AI technology evolves more towards autonomous intelligence, will that legal framework hold up?
As an example, the Patent Act provides that “whoever without authority makes, uses, offers to sell, or sells any patented invention, within the United States . . . during the term of the patent therefor, infringes the patent.” What if AI is the actor that makes or uses a patented invention independently? If “whoever” in the Patent Act is construed to only encompass humans, companies, and governmental entities, then it is possible the AI will be deemed to not infringe. On the other hand, if a human or company then sells or offers to sell the output from the AI’s manufacture or use of the patented invention, the human or company may be liable for infringement. The analysis will become only more difficult in situations where the human or company may not know whether the AI used or made a patented invention.
As another example, the Patent Act provides a medical procedure exception that no damages may be obtained for a “medical practitioner’s performance of a medical activity that constitutes an infringement.” The rationale is that Congress does not want doctors to worry about infringing a patent when they are trying to help patients. That exception extends to the “medical practitioner” and the “related health care entity with respect to such medical activity.” But does such an exception extend to contexts such as robotic-assisted surgeries or virtual nursing assistance, and if so, how far? Will the liability standard for AI developers who create robots for robotic-assisted surgeries or virtual nursing be the same way as for medical device manufacturers whose devices are used in medical procedures that potentially infringe a patent? Should the liability be the same when the robots can be “actors” in a way that typical medical devices cannot?
The current IP legal framework does not address many issues that are likely to arise as AI becomes increasingly prevalent. AI is, therefore, likely to have a big impact on the legal framework for IP in the upcoming years.
For more information, please contact Lauren J.F. Barta or Tim Grimsrud.
[1] For source information, please contact the authors.
