AI Inventors? Not So Fast.

Sooner than I expected, there’s an update to this blog post from way on back in June. That took as its subject the effect of AI systems on patent inventorship, which not too many years ago would have sounded like speculation on whether or not hypothetical Martians could also get in on some of that sweet patent action. But it’s become more of a real issue, and (like most anything else to do with AI) is becoming rather more real all the time.

In that post, I mentioned the attempts by a firm called Imagination Engines to have an AI system named as an inventor on its patents, and how that approach had been thus far rejected almost everywhere it had been tried (South Africa being the exception). Well, the US Court of Appeals for the Federal Circuit (the highest venue for patent cases short of the Supreme Court) has been considering the case here in the US, and ruled last Friday in Thaler v. Vidal that inventors must indeed be human beings. Stephen Thaler is the Imagination Engines guy, and he had two patents (for a package design and a lighting beacon) rejected when he applied for them with his company’s AI software listed as sole inventor. He appealed that decision to the Federal District Court, and after losing there took his appeal to the CAFC, where he has now lost again.

The three-judge panel held that nothing in the statute law on inventorship can be construed as allowing for anything other than a human inventor. The Patent Act says that an inventor is the individual or individuals who discovered or invented the subject of the patent, and the CAFC went back to another Supreme Court case that had occasion to say that “individual” commonly refers to a human being (as opposed to a corporation or a sovereign state, for example). Furthermore, the law makes references such as “himself or herself” and the like, which are used to refer to human beings rather than to software or machines.

The CAFC completely dodged the more philosophical questions involved, and I can’t blamed them. This is purely a ruling on statutory interpretation, not on what’s theoretically possible. So basically, they are indicating that if you want patent law to allow for AI inventorship, then Congress is going to have to pass a new law explicitly saying that, because it is simply not in the law as written. Now, if they ever do – and I’m not going to wait on that – then there will surely be challenges based on those more philosophical questions – what really is an invention anyway, what is creativity and is it something that software can demonstrate, and if so is that just a reflection of the creativity of the programmers and operators of the system themselves, and so on and so on. The courts will deal with those questions if and when they arise, but the CAFC is determined to make clear that they have not yet arisen because of the laws we have right now.

Thaler himself is saying that he wants to appeal to the Supreme Court, but I think his chances of being “granted cert” there are extremely small. The Supreme Court really does not like dealing with patent cases, preferring to defer to the more specialized CAFC whenever possible, and typically only steps in to clear major logjams. This isn’t one – or at least not yet. I absolutely expect them to turn down the petition – and if anyone were asked to explain why, the answer would be that the USPTO, the District Court, and the Court of Appeals for the Federal Circuit did not, in fact, manage to all be wrong in the same way on this case, and that US statute law is clear on the matter. And if you read Supreme Court decisions with any regularity (pretty interesting stuff, not going to lie) you’ll know that in a great many cases the rulings are along the same lines: “If Congress wanted to say XYZ, they could have – but they didn’t. If they want to say it now, they most certainly can pass a law to that effect. But since they have not, we therefore rule that. . .” This particular tent peg doesn’t look (yet) like it needs to be pounded into the soil any more.