I grew up thinking about how awesome it would be to hang out on the holodeck aboard the Starship Enterprise and how terrifying it would be to be relentlessly chased by a stone-cold Terminator. But back in the day, artificial intelligence (AI) was just the faraway stuff of dreams.  

Today, AI is a reality that promises amazing, super-accelerated advancements in technology (while also creating the potential for nightmares that no one wants floating around in their head late at night!).  

Isaac Asimov brilliantly contemplated his Three Laws of Robotics for keeping robotic artificial intelligence in check more than a half-century ago – that a robot shall not injure humans, shall obey its master, and shall protect its own existence as long as it doesn’t violate the first two laws. 

But real-world laws surrounding artificial intelligence are way behind the times. In January 2019, the director of the USPTO finally acknowledged that policy makers need to consider:

  • Will the legal concepts of ‘inventor’ or ‘author’ fundamentally change due to AI?
  • Who retains title to an improvement developed by a machine — the original programmer, machine owner or even the machine itself?
  • Should using copyrighted works to “train” AI systems constitute fair use or some other exception?
  • How do we assess patentability as well as the risk of bias when there is little transparency regarding how algorithms are trained and function?
  • How will firms, both large and small, protect AI-related inventions?

The USPTO is also currently requesting comments on artificial intelligence AI inventions. 

Non-Human Ownership Issues Are Nothing New

The question of assigning ownership to non-humans has been making headlines for years. 

For example, do you remember the monkey selfie that went around a few years ago? The photo became famous, primarily because of the disputes surrounding it. The travel photographer that was present when the monkey took the photo said that he has a valid copyright claim to the photo because he “engineered the situation that resulted in the pictures.” Meanwhile, PETA said the monkey should be assigned the copyright and Wikimedia Commons said the photo is in the public domain, as a non-human creator cannot hold a copyright. The photographer claimed that the dispute cost him considerable money. In the end, PETA and the photographer reached a settlement. The courts decided that animals don’t have standing and can’t bring a lawsuit. 

Similar to how animals are neither inventors nor copyright holders, at least under the U.S. law, currently machines are not inventors or copyright holders. But could that be changing quickly?

As widely reported, a team led by the University of Surrey is pushing the definition of “inventor” by pursuing patent applications that cover ideas autonomously developed by AI (without a human named as inventor). The named AI “inventor” instead is referred to as DABUS, short for Device for the Autonomous Bootstrapping of Unified Sentience. According to reports, one patent application is directed to a beverage container and another is directed to a device to help attract attention that may be used in the context of search and rescue.

It’s All in the “Who”

The USPTO (and any U.S. court taking up the issue) is likely to continue finding that presently only humans are entitled to be the inventors of a patent. There is no statute that says machines are not entitled to patents, but 35 U.S.C. § 102 explicitly states a “person shall be entitled to a patent” if statutory requirements are met, and Congress has not conveyed any intent to allow machines to be inventors. Additionally, there are current requirements for inventors to sign declarations saying that they were the ones who conceived their idea and that they didn’t take their idea from someone else, under penalty of jail time (for making a false statement).

Substantial changes to existing laws would be required to accommodate machines being innovators or inventors.

Related: Pitfalls to Avoid When Collaborating on Intellectual Property 

Potential Benefits of AI-Created Patents

Incentivizing AI innovation by granting a patent for a limited time could come with some potential benefits. 

Those in favor argue that patent protection is necessary, otherwise there will be insufficient incentive to realize ideas that may solve the world’s biggest health or environmental issues. For example, a pharmaceutical derived from AI is an enormously expensive pursuit that may not ever occur without patent protection.

1984 All Over Again?

But do the potential benefits outweigh the potential costs? Could losing the human aspect of invention lead to a scenario straight out of a science fiction novel? 

It’s just a given that large companies have more access to AI technology than small companies. Giving patents to AI, then, gives unfettered, larger companies an innovation edge.

Imagine this nightmarish scenario: AI mines through virtually all existing patents, “watches” for new published patent applications, and on an ongoing basis creates new patent claims (i.e., potential inventions) to fill in holes not covered by existing patents. AI maybe (or maybe not) models the new claims to assess whether the claims are, in fact, useful and then AI automatically drafts patent applications that are ready for filing. 

Already, AI helps draft patent applications via basic information that a human user feeds into a program. A little fine tuning and suddenly you have autonomous machines coming up with new ideas without even a seed of anything human, just overwhelming the patent office with new inventions, all being funded by large companies. And, unlike a human inventor, AI never gets tired, never burns out, and never runs out of steam.

Related: What Lawyers and Clients Need to Know About the Use of Artificial Intelligence

In this scenario, it’s easy to imagine well-funded companies quickly covering entire technological patent landscapes in a way an army of engineers and patent attorneys cannot do. It is also possible an entire country may unleash a race to the bottom by being the first to deregulate patent laws and allow non-human inventors.

Striking the Right Balance

Special laws and rules might need to be put in place to throttle AI-related patent filings. For example, requiring that AI-developed innovation actually be put into commercial practice before patent rights are granted may strike an appropriate balance. Other measures to address the potential for an overwhelming number of AI-as-an-inventor patent filings may include compulsory licensing and per-company limits on the number of AI-as-an-inventor patent filings. If you have thoughts, send them to the USPTO at AIPartnership@uspto.gov by November 8, 2019. 

AI promises enormous potential benefits, but requires amendments to laws and rules that will protect innovators and inventors — especially those with limited resources — while still taking advantage of all that AI offers.