Humans Only - US Federal Court Rules that AI Cannot Own Copyright in Created Works

This blog has on many occasions discussed the difficult world that IP can be for both AI creations and inventions (e.g. more here, here and here), with the landscape looking more and more 'hostile' towards AI creations and inventorship. This hasn't, however, put a damper on many who will keep trying to push the envelope, and test the limits of IP laws across the world. Such an individual is the prolific inventor, Stephen Thaler, who has been on a crusade to enable AI to own many IP rights, having failed on many occasions along the way. His saga has since continued, with the US Federal Court being the next court to consider AI authorship this month.

The case of Stephen Thaler v Shira Perlmutter, Register of Copyrights an Director of the US Copyright Office concerned a system called "Creativity Machine", which Mr Thaler owns. Creativity Machine has, according to Mr Thaler, generated a piece of visual art of its own accord, which he then sought to register on the system's behalf. This application was rejected by the US Copyright Office for a lack of human authorship (see link above). Mr Thaler did not accept this outcome and appealed the decision, which landed on the desk of Judge Howell. 

Both parties sought summary judgment on the matter, for which Judge Howell handed down their judgment. At its core the matter concerned "whether a work generated entirely by an artificial system absent human involvement should be eligible for copyright".

Judge Howell first noted that, pursuant to the US Copyright Act 1976, copyright protection attaches immediately upon the creation of “original works of authorship fixed in any tangible medium of expression", provided those works meet certain requirements. 

The judge also noted that "copyright law has proven malleable enough to cover works created with or involving technologies developed long after traditional media of writings memorialized on paper", which gives the Act scope to broadly adapt to changing technologies and methods of creation. 

However, despite this adaptability, at its core copyright still requires human creativity, even if that creativity is "channeled through new tools or into new media". The judge determined that "[h]uman involvement in, and ultimate creative control over, the work at issue was key to the conclusion that the new type of work fell within the bounds of copyright" (referring to the decision in Burrow-Giles Lithographic Co. v Sarony). 

To really drive the point home, Judge Howell states that "[c]opyright has never stretched so far, however, as to protect works generated by new forms of technology operating absent any guiding human hand, as plaintiff urges here. Human authorship is a bedrock requirement of copyright". The important point here is that the decision requires that a human guides the creation of a work by AI systems, so an absence of human intervention can lead to the works falling outside of copyright protection. 

As has been mentioned in earlier decisions, this position is corroborated by the text in the Act itself, which notes that copyright attaches to "original works of authorship fixed in any tangible medium of expression, now known or later developed, from which they can be perceived, reproduced, or otherwise communicated, either directly or with the aid of a machine or device". The fixation of the work must also be done "by or under the authority of the author". 

While the term 'author' is not defined in the Act, the judge quickly decided that the Act requires a copyrightable work to have an originator with the capacity for intellectual, creative, or artistic labour who specifically is a human. Additionally, "[t]he act of human creation—and how to best encourage human individuals to engage in that creation, and thereby promote science and the useful arts—was thus central to American copyright from its very inception".

The requirement for human authorship has also been followed and kept as copyright laws have evolved, with the Supreme Court consistently ruling that it is a crucial requirement for copyright protection (e.g. in Sarony above, Mazer v Stein and Goldstein v California). 

The courts have also consistently rejected copyright ownership by non-human authors. This has included divine beings (Urantia Found. v Kristen Maaherra), gardens (Kelley v Chicago Park District) and monkeys (Naruto v Slater). 

Judge Howell nevertheless acknowledges that in the future the courts and/or legislatures will have to grapple with "how much human input is necessary to qualify the user of an AI  system as an “author” of a generated work, the scope of the protection obtained over the resultant image, how to assess the originality of AI-generated works where the systems may have been trained on unknown pre-existing works, how copyright might best be used to incentivize creative works involving AI, and more". 

Judge Howell concluded that the US Copyright Office acted properly in denying copyright registration for a work created absent any human involvement. 

The decision is one in a long string which have denied AI authorship and inventorship, and it looks unlikely that this will change bar legislative amendments to existing copyright regimes. The current, admittedly nearly archaic legislative frameworks have not, and frankly could not have, conceived of a world where computers or AI systems could create as they do currently. This leaves legislative action as imperative, lest we want the law to forever lag behind technology. 

Just Can't Catch a Break - US Court of Appeals Decides That AI Cannot be an Inventor of Patents

Artificial intelligence's road to recognition as a legal entity keeps hitting obstacle after obstacle, and as things stand it looks like right's in AI created works or inventions will be questionable without proper legislative intervention. Despite the continuous stumbling blocks, it seems that some inventors are very interested in continuing the fight, so yet another important decision has been reached on AI inventorship of patents in the US. The Court of Appeals recently handed down its decision on the matter, which continues on the near never-ending line of decisions across the globe that have been discussed on this blog before many times (for example here, here and here). 

The case of Stephen Thaler v Katherine Vidal concerned the DABUS AI system, which will be very familiar to the readers of this blog by now. DABUS was created by Mr Thaler and set out to create a number of inventions, and subsequently Mr Thaler applied for patents for some of these inventions (at issue specifically applications 16/524,350 and 16/524,532). In these applications Mr Thaler listed DABUS as the sole inventor. Mr Thaler also handled all of the filing formalities for DABUS, including a number of documents setting out why it is the inventor of the inventions. The USPTO rejected both applications due to them lacking a valid inventor and therefore being incomplete. The District Court, on appeal, found that "an 'inventor' under the Patent Act must be an 'individual' and the plain meaning of 'individual' as used in the statute is a natural person", and Mr Thaler therefore appealed to the Court of Appeals. 

Justice Stark noted that the sole issue in the case is whether an AI software system can be an "inventor" under the Patent Act. 

The Patent Act expressly sets out that inventors are 'individuals', which is means "the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention". The key here is the word individual, which Justice Stark highlighted to meaning "...a human being" (as decided by the Supreme Court in Mohamad v Palestinian Authority). Unless defined otherwise by statute this remains the definition, as is the case under the Patent Act. 

When not inventing, AI likes to take it easy with
a little light gardening in the evenings

Despite the wording 'whoever' being used in the Patent Act, the fact remains that these provisions are still subject to the above, i.e. that an inventor be a human being. Mr Thaler also noted that AI software programs should qualify as 'inventors' since otherwise patentability would depend on "the manner in which the invention was made". Justice Stark rejected this position, specifying that this isn't about inventorship, and also that that inventions may still be non-obvious even if they are discovered during 'routine' testing or experimentation. 

Existing precedent supports this position, with, for example, the decision in the University of Utah v Max-Planck setting out that "inventors must be natural persons and cannot be corporations or sovereigns", and in Beech Aircraft Corp v EDO Corp where it was decided that "only natural persons can be 'inventors'". Justice Stark noted that, while they deal with a different question (i.e. corporate ownership), the cases nonetheless support the position that the plain meaning of 'inventor' in the Patent Act is limited to natural persons.

Justice Stark then touched on Mr Thaler's various arguments to round out their decision. Mr Thaler had argued that inventions generated by AI should be patentable in order to encourage innovation and public disclosure. Justice Stark determined this to be purely speculative and lacking in basis under the Patent Act. Mr Thaler also argued constitutional avoidance where permitting AI programs to be inventors would support the constitutional purpose of patents "to promote the progress of science and the useful arts" and that not recognizing AI as an inventor undermines such progress, raising potential constitutional concerns that we should avoid. Justice Stark also readily rejected this position, setting out that the lack of AI inventorship would be unconstitutional, since Congress has decided to legislative limiting inventorship to human beings. 

The Court of Appeals therefore rejected Mr Thaler's appeal. 

The decision was, by no means, a surprise and continues the trajectory of rejection that AI inventorship has faced over the last several years (despite the South African CIPC registering DABUS' inventions as patents, being the only jurisdiction to do so so far). It remains to be seen whether Mr Thaler will want to try and take the matter to the Supreme Court, but it would seem unlikely to be picked up due to the sheer clarity of the matter in the US. A future involving AI inventorship would clearly require further steps from legislatures across the board, but currently there seems to be very little appetite to do so.

The EPO Are Not 'Board' of AI Yet - EPO Board of Appeal Weighs in on Whether Artificial Intelligence Can Be an Inventor

The progression of AI in the world of IP rights keeps trundling along, with many national and international bodies having already weighed in on whether AI can be an inventor for the purpose of patent law or not (discussed on this blog before). The answer seems to have been a resounding "no", however, following an earlier decision by the EPO, the EPO Board of Appeal has now issued its decision in the ever-continuing DABUS saga, which does shed some further light on the matter and does show that the current legislative framework does seem robust enough on the matter of AI (though this writer would think that specific legislative updates are still a must).

The case of Designation of inventor/DABUS (case J 0008/20) concerned two patent applications filed at the EPO (namely EP18275163 and EP18275174). The applications didn't list an inventor for the patents, but, following a clarification by the applicant, Stephen Thaler, the inventor was noted to be "DABUS". DABUS is an AI created by Mr Thaler. The EPO Receiving Section rejected both applications on the grounds that the designated inventor, DABUS, didn't meet the requirements for an inventor, that being a need for them to be a 'natural person'. Mr Thaler subsequently appealed both decisions to the Board of Appeal.

The appeal was brought on three separate grounds: (i) whether an applicant can designate an entity which is not a natural person as the inventor under Article 81 of the European Patent Convention; (ii) to comply with the EPC is it enough for an applicant to file any declaration irrespective of its content, or does the latter need to satisfy specific requirements; and (iii) whether and to what extent the EPO can examine and object to statements filed under Article 81.

To set the scene, Article 81 requires that an inventor be designated for all patent applications, and any deficient designation of an inventor can be amended under Rule 21 of the EPC. Further, Article 61 of the EPC notes that "[t]he right to a European patent shall belong to the inventor or his successor in title" (the latter being a legal successor in the title of the rights), and the rights of any employee, if they are the inventor, will be determined by the national legislation where they are employed. 

The Board of Appeal first dealt with the main request in the matter, i.e. whether an AI can be an inventor of a patent. The Board firmly rejected this point, as they determined that "[u]nder the EPC the designated inventor has to be a person with legal capacity", so adopting a simpler, ordinary meaning approach in deciding what an 'inventor' is. What the provisions around designating an inventor seek to do are to confer and protect the rights of the inventor, facilitate the enforcement of potential compensation claims provided under domestic law, and identify a legal basis for entitlement to the application. The Board clearly set out that "[d]esignating a machine without legal capacity can serve neither of these purposes". 

The Board also noted that there is no practice or agreement that would allow for the Board to overcome this specific language. The appellant also advanced an argument of fairness allowing for AI to be an inventor of a patent; however, the Board immediately rejected this argument as well. 

The Board therefore confirmed that the Receiving Section's decision was correct in raising an objection to the applications. 

In summarizing its position, the Board decided that "…the main request does not comply with the EPC, because a machine is not an inventor within the meaning of the EPC. For this reason alone it is not allowable". 

The Board then turned to the auxiliary request, i.e. whether the first sentence of Article 81 does not apply where the application does not relate to a human-made invention. The Board swiftly sided with the request and agreed that the part of Article 81 would not apply in that instance, since the provisions were drafted to confer specific rights on the inventor, so where no human inventor can be identified, then Article 81 does not apply. 

The appellant had provided a statement with this request noting that they had derived the right to the European patent as owner and creator of the machine. The Board disagreed and stated that the statement didn't bring them within the remit of Article 60, since there is no legal situation or transaction which would have made him successor in title of an inventor. 

The Board did consider two referrals that were requested to be made to the Enlarged Board, however, the referrals were rejected. 

In discussing why no referral was needed, the Board turned to matters surrounding Article 52 of the EPC, which specifies that an invention is one which is novel, industrially applicable and involves an inventive step is patentable. The appellant argued that the provision only relates to human-made inventions, and the Board agreed with this position. How a particular invention was made doesn't play a part in the patent system, so therefore it is possible that AI-generated inventions too are patentable under Article 52. However, this position can't co-exist with Article 60, so you would have inventions that are patentable under Article 52, but for which no rights are provided under Article 60. 

The Board also considered the requirement to file a statement of origin of the right to a European patent where the inventor and the applicant differ. The Board noted that this is only a formal requirement, and only informs the public of the origin of the right, but that it would be disproportionate to deny protection to patentable subject-matter for failing to fulfil such a formal requirement. 

Additionally, they noted that, as the lawmakers only had human-made inventions in mind when drafting Article 60, so it is possible that no statement on the origin of the right is required where the application concerns an invention developed by a machine. 

The crux of why Mr Thaler's applications have failed in light of the above is that there is nothing preventing him from listing himself as the inventor in relation to both applications. The Board highlighted that there is no case law "…which would prevent the user or the owner of a device involved in an inventive activity to designate himself as inventor under European patent law".

The decision is yet another among many that highlight the issues of AI inventorship and why there is simply a need for evolution in the law to accommodate AI inventions somehow. However, as noted by the Board, the creator of the AI could be listed as the inventor, bypassing this issue, however, it seems that Mr Thaler was deadset on setting a new precedent and allowing for direct ownership by the AI systems. This presents the problem of enforcement and/or remuneration, and undoubtedly these would be handled by the creator themselves, rendering this problem almost moot in this writer's view. We will see if the case progresses to the Enlarged Board, but it seems that Mr Thaler is in no way done with the DABUS saga as things stand. 

Yet Another Rejection - US Copyright Office Rejects Registration of Copyright in AI Created Artwork

How works created by artificial intelligence, or AI, are treated by various jurisdictions has been a hot topic recently, including in this very blog (for example here and here). It seems that, despite a briefly celebrated win in Australia, the technology is seeing setback after setback on whether AI-created works will be covered by IP and this seems more unlikely every day. The US Copyright Office was the latest blow in the battle for IP protection of AI-created works, which leaves the technology in a bit of a bind, however, it featured a familiar individual yet again. 

The US Copyright Office's Review Board recently released a decision on the copyright protection of a work, A Recent Entrance to Paradise, created by an AI program called Creativity Machine, developed by Steven Thaler. While the Creativity Machine was the author of the work, Mr Thaler was listed as the claimant alongside it as its owner. The work was autonomously created by the Creativity Machine, but Mr Thaler wanted to register the work as a work-for-hire to the owner of the program, himself. 

The US Copyright Office initially rejected the registration of the copyright in the work in 2019 as it "lacks the human authorship necessary to support a copyright claim", which Mr Thaler subsequently appealed this decision arguing that the "human authorship requirement is unconstitutional and unsupported by either statute or case law". Despite this, the Copyright Office rejected the registration on the same basis as before and noted that Mr Thaler had not provided any evidence of sufficient creative input or intervention by a human author in the work and that the Copyright Office won't abandon its longstanding interpretation of this matter which is based on precedent from the courts, including the Supreme Court. 

Being ever-persistent, Mr Thaler then yet again appealed, arguing in a familiar fashion that the Office’s human authorship requirement is unconstitutional and unsupported by case law. Mr Thaler also argued that there is a public policy piece to this where the Copyright Office 'should' register copyright in machine-created works because doing so would "further the underlying goals of copyright law, including the constitutional rationale for copyright protection". Mr Thaler also again rejected the Copyright Office's assertions of supporting case law. 

At the beginning of its decision, the Review Board accepted that the works were indeed created by AI without any creative contribution from a human actor. However, the Board noted that copyright only protects "the fruits of intellectual labor that are founded in the creative powers of the [human] mind". Mr Thaler was therefore tasked with either proving that the work was created by a human, which it hadn't been, or somehow convincing the Board to depart from long-established precedent.

The Review Board then moved on to discussing the precedent relevant to the matter at hand. 

Paragraph 306 of the US Copyright Compendium sets out that "the Office will refuse to register a claim if it determines that a human being did not create the work".  As such, human authorship is a prerequisite to copyright protection in the US. 

Similarly, 17 USC 102(a) affords copyright protection to "original works of authorship", and, while the phrase is very broad in its definition, it still requires human authorship. This has been supported by US Supreme Court decisions in Burrow-Giles Lithographic Co. v Sarony and Mazer v Stein where the Supreme Court required human authorship for copyright protection. In addition to the highest court in the US, the lower courts have also followed these decisions. 

Although the courts haven't directly decided on AI authorship in relation to copyright just yet, Mr Thaler has featured in a recent decision in the case of Thaler v Hirshfeld (the very same individual as in this decision) where the court decided that, under the Patent Act, an ‘inventor’ must be a natural person and cannot be an AI. 

The Board also discussed a recent report by the USPTO who discussed IP issues around AI. In this report, the USPTO noted that "existing law does not permit a non-human to be an author [and] this should remain the law". 

The Board also briefly touched on Mr Thaler's secondary argument, which was that AI can be an author under the law because the work-made-for-hire doctrine allows for non-human, artificial persons such as companies to be authors. The Board rejected this argument since the work was not made for hire. This was due to the requirement for such a work to be prepared by e.g. an employee or one or more parties who agree that the work is one for hire. No such agreement was in place between Mr Thaler and the Creativity Machine. Even outside of this, the doctrine merely addresses ownership and not copright protection, so human authorship is, yet again, required. 

The Board, therefore, rejected the appeal and refused to register the work. 

The decision by the Board is by no means surprising, as the status of AI-created works seems to be more and more established as the years go by as a no-go. While copyright protection could be possible in some jurisdictions, as discussed in the above blog articles in more detail, it seems highly unlikely to be possible in the US without legislative intervention. As AI develops the law will have to address this and potentially grant AI-creation some level of protection, however, this will take some time to come true. 

An Unusual Inventor - Australian Federal Court Gives Green Light for AI Inventors of Patents

Having spoken about whether artificial intelligence could be deemed to be an inventor was something that was discussed on this blog over a year ago, with both the UK, the US and the EU, currently, not allowing for AI to be considered as an 'inventor' and therefore any inventions created by the AI would not be patentable. While this position seems to remain in those jurisdiction, an interesting development has emerged in Australia which seems to go counter to what the above jurisdictions have decided. 

The case Thaler v Commissioner of Patents concerned a patent application for a patent by Stephen Thaler (application VID108/2021) for a patent for two inventions, created by DABUS (Mr Thaler's AI system), for an improved beverage container and a flashing beacon to be used for emergencies. The case focused on whether the AI system could be an inventor under the Patents Act 1990. The Deputy Commissioner of Patents initially rejected the application as, in their view, the application didn't comply with specific requirements and that under s. 15 of the Patents Act an AI could not be treated as an inventor. Mr Thaler then sought a judicial review of the decision, which ended up in the Federal Court.

The Court first discussed the issue generally, noting that: (i) there is no specific provision in the Patents Act that expressly refutes the proposition that an artificial intelligence system can be an inventor; (ii) there is no specific aspect of patent law, unlike copyright law involving the requirement for a human author or the existence of moral rights, that would drive a construction of the Act as excluding non-human inventors; (iii) the word "inventor" has not been expressly defined and under its ordinary meaning it can include any 'agent' which invents, including AI; (iv) the definition of an "inventor" can be seen to potentially evolve and change with the times, potentially to include non-human authors; and (v) the object of the Act is to promote the economic wellbeing of the country, which should inform the construction of the Act. 

To include AI as an inventor would, according to the Court, help in promoting innovation and the promotion of economic wellbeing in Australia. But the problem of who the ultimate legal owner of the rights still remains. This would indeed promote the objective of s. 2A of the Act, which sets out the object of the Act as discussed above. 

The Court highlighted the need for a human applicant, on behalf of the AI system, for any inventions that these systems might come up with, who would then also have the rights to any patents granted for that AI system. 

The Court also discussed the notion of the "inventive step" under s. 18 of the Act, which is a legal requirement for patentability. Both s. 18 and 7, which elaborates more on the meaning of the "inventive step", don't require an inventor per say, nor that such an inventor, if even required, would be legal person. The Court concluded that, for there to be an inventive step, a legal person isn't required as the inventor at all, even though the Act does refer to "mental acts" and "thoughts".

The Court then moved onto discussing the various dictionary definitions for what an "inventor" is. This was quickly dismissed as a grounds to limiting inventors to legal persons, as the Court noted that the definitions, as set out, have moved on from the historical meanings given to them and can't be limited in the same way. 

The next issue was who the patent would be granted to under s. 15 of the Act (but focusing on still who could be classed as an "inventor"). The section includes four different classes of person to whom a patent can be granted. 

The first is if the inventor is a person. The Court quickly determined that, as DABUS is not a person that the section won't apply in this instance, but also discussed that this does not demonstrate that the concept of a "person" would be different to that of an "inventor". An AI system could indeed be the inventor, but not able to be granted a patent as they don't fulfill the requirements of s. 15. 

The second one is a person who would, on the grant of a patent for the invention, be entitled to have the patent assigned to them. This will be discussed in more detail following the other classes. 

The third one is a person who derives title to the invention from the inventor or a person mentioned in the second class. Again, this will be discussed further in the context of the second class. 

Finally, the fourth one is a person who is the legal representative of a deceased person in one of the earlier named classes, which, unsurprisingly, the Court determined to not apply here.

Returning to the question of the second and third classes, the Court first highlighted that Mr Thaler could indeed fall under the second class as the programmer and operator of DABUS, through which he could acquire title in any inventions that may be granted patents over. The title in the patent flows through to Mr Thaler automatically, so wouldn't require the assignment of any rights by the inventor (here DABUS) to Mr Thaler. Additionally, the Court determined that s. 15 doesn't require an inventor at all, but only requires that an applicant is entitled to have a patent assigned to them. 

Also, Mr Thaler, according to the Court, would fall under the third class as he, on the face of it, he has derived title to the invention from DABUS. As the AI system cannot legally assign any inventions to Mr Thaler, however, the language of s. 15 does allow for one to derive rights in an invention even outside of legal assignment of those rights. All of Mr Thaler's rights in any invention become his by virtue of his ownership and operation of the inventor, DABUS. 

The Court succinctly summarised the matter as "generally, on a fair reading of ss. 15(1)(b) and 15(1)(c), a patent can be granted to a legal person for an invention with an artificial intelligence system or device as the inventor".

Ultimately what the case focused on is whether a valid patent application has been made, rather than who will own any patent that might be granted in the future. 

What the Court determined was that "...an inventor as recognised under the Act can be an artificial intelligence system or device. But such a non-human inventor can neither be an applicant for a patent nor a grantee of a patent". This still leaves the ownership of a patent in the air, but, focusing on what was discussed by the Court, it is more than likely that ownership in any AI inventions would automatically pass to the owner and operator of that AI system. 

The Australian Court's approach to this question has hugely departed from the viewpoints of that of the US, UK and the EU, where AI systems have been rejected as potential inventors. It will be very interesting to see the law change and evolve in this in the years to come, in particular any decisions on the granting of a patent to the owner of an AI system, but so far at least Australia seems to be on the vanguard of allowing for AI systems to potentially push the envelope on innovation and to protect those innovations in the process.