To quote from the venerable protocol droid, C-3PO: ‘Thank the maker!’ Or in this case let’s thank the UK’s Court of Appeal in London, which has ruled that ‘AI systems cannot own or transfer patent rights’, as explains a legal update from law firm Pinsent Masons.
Pinsent Masons lawyers, Sarah Taylor and Mark Marfé, highlighted the recent case of Stephen Thaler, who has been seeking to patent inventions that he claims were derived from an AI system called ‘DABUS‘.
‘The Court of Appeal found that only a person can be an inventor, and that as Dr Thaler accepts that he is not the inventor, he has no entitlement to the patent,’ the lawyers explained.
It was also DABUS, which is a bunch of computer code written by good old-fashioned humans, that succeeded in South Africa in being recognised as the owner of a patent related to a ‘food container based on fractal geometry’.
The argument is that as DABUS comes up with the many varieties of solution to a design problem by itself, then it should be regarded as the inventor of whatever it makes, and therefore holds the patent. That patent is then presumably transferred over to the owner of DABUS, as its lines of computer code don’t have much use for the money that can be made from inventions.
But, this is a bit of a stretch, and on that basis your toaster is the ‘inventor’ of your toast, given that this human-made device never produces exactly the same outcome and is operating autonomously, (at least once you stick the sliced bread in and pull down the lever). I.e. the toaster is creating something unique.
A toaster is less complex in terms of its ‘programming’ than DABUS and other problem solving systems, but it’s still on the same plane of existence. DABUS is just computer language that someone wrote, it’s a manufactured ‘device’.
The problem with the idea of ‘AI’ as an inventor is that it appeals to the red herring theory that because machine learning technology can mimic human behaviour it therefore has the qualities of human identity and thought. But, this is just anthropomorphism for the 2020s. As mentioned, if you follow that logic then your toaster was the ‘inventor’ of your toast this morning, as it created it, and that output – if examined in detail – is unique to that toaster. (Although, admittedly, it would be hard to patent your toast….unless it was very, very special.)
Perhaps the root of the problem is that if you press the point that a piece of software is the inventor, then you are digging a hole for yourself when you come to ask to own the patent for that invention. It would have been a lot simpler – one presumes – for someone to just do their research, use a bunch of software to help with that, and then claim the patent for what you invented – as most people do. But, by persisting with the idea that ‘it was the machine that did it’ you put yourself in an impossible position.
One might say that the hubris of wanting your AI system to be considered as like a human, in the end leads to one’s own downfall as a potential patent holder.
But, this is not the end of the story. As the Pinsent Masons lawyers noted: ‘However, there was a split in the judgment [at the Court of Appeal], with two of the three ruling judges – both leading authorities on UK patent law – in disagreement on the topic.
‘This means that the UK Supreme Court is likely to be asked to rule on the core question of whether an AI system can be named as an inventor under UK law, and it could further expedite a change to UK legislation.’
So, this isn’t over yet. But, hopefully the wise judges of the UK’s Supreme Court will see the nuttiness of the whole endeavour.
Here is some more insight from the Pinsents team. The full briefing can be found on their website.
‘Dr Thaler has been arguing across several jurisdictions that the owner of AI systems should be the default owner of patents for inventions derived from those systems, and that it should be possible to name those AI systems as inventors on patent applications. The High Court ruled against Dr Thaler last September, upholding the position taken by a senior official at the UK Intellectual Property Office (UKIPO) in the process. The European Patent Office (EPO) also rejected Dr Thaler’s arguments last year, as did a district court in the US state of Virginia more recently. However, a parallel case about applications for the same inventions was decided in Dr Thaler’s favour in Australia, though it is now being appealed. An application corresponding to one of Dr Thaler’s UK patent applications has also been granted in South Africa.
In the High Court ruling, Mr Justice Marcus Smith held that the Patents Act 1977 provides that a person making a patent application must be a ‘person’ with legal personality, whether a human or corporation, and that a patent can only be granted to such a ‘person’ with legal personality.
He determined that, because the inventor is by default the person entitled to the patent rights, it followed that existing legislation requires the ‘inventor’ to be a person with legal personality. The judge further held that, because patent rights are property rights, a machine is incapable in law of holding and transferring patent rights since it lacks the legal personality necessary to assign the rights to property or even hold those rights in the first place.’
Interesting case. I don’t know the reasoning, but would say that arguing for an AI to be treated as a human is one approach, arguing for it to be treated as a legal entity is another. We do that with companies, which are proper abstractions in themselves, so are AI’s. This approach then calls for regulating ownership of implementations of software, algorithms and AI models. This, of course, is not an easy topic, because we’d be looking at ownership of the state of a part of computer memory on a specific machine, which is (i) not recognized as a good in most legal systems and (ii) dynamic in itself.
There´s harmonization to be done in the IPR arena – e.g. for copyright in the UK see 9(3) Copyright, Designs and Patents Act 1988. (don´t know if there´s any court decisions related to this though)
“In the case of a literary, dramatic, musical or artistic work which is computergenerated, the author shall be taken to be the person by whom the arrangements
necessary for the creation of the work are undertaken.”
I don´t know if there´s similar examples in other jurisdictions (can´t recall having seen this explicit regulation in other places). Wihtout going into details, determining ownership to computergenerated (copyright) works under Swedish law would not be carried out in the same way.
Addition: How the originality assessment under UK law impact the “computergenerated” works regulation is beyond my knowledge so I´ll stay silent on that topic 🙂