Can Copyright Survive Artificial Intelligence?

February 8, 2018

At present, English law is one of the few bodies of jurisprudence to have addressed the question of copyright in the output of a machine. By the Copyright Designs and Patents Act 1988, s 9(3), the author of a computer generated literary, dramatic, musical or artistic work is deemed to be the person by whom the arrangements for the creation of a work were undertaken. This provision has worked perfectly in the subsequent quarter century, but may now be reaching its limit. 

The section assumes that whenever a computer generates a work there has to have been a person who made the arrangements enabling this to happen. That assumption was valid in 1988 – 9 years before IBM’s Deep Blue beat Gary Kasparov at chess. But AI has recently moved beyond the brute force triumph of mastering all possible moves in a game of chess to beat a grandmaster, to AlphaGo teaching itself in just three days to play Go far better than any human simply by testing possible outcomes based purely on the rules. An AI-compiled short novel reached the short-list for a Japanese literary prize (albeit it did not win: lack of character development was particularly criticised). Two students have developed a machine-learning program, GANGogh, able to produce original artworks of a standard that lay audiences might well not distinguish from works of a significant human artist. And the ability of computers to write music – such as ‘Mozart’s 42nd Symphony’– had been well established even before machine learning techniques were developed. In 2016, Sony’s Flow Machine software generated its first pop song, designed in the style of The Beatles. 

Once AI is unleashed on any creative problem, we can anticipate truly original compositions. This brings two questions into sharp relief. Should copyright protection be given to works that were independently created by a computer, without input from any human ‘author’ beyond developing code which can teach itself new skills? And if computers can produce works equivalent in originality to much workaday literary, dramatic and musical material, does it make sense to confer copyright on such work even if it happens to be produced by a human? 

Both questions depend upon the rationale for conferring monopoly rights as a reward for creative work in the first place. Two unrelated justifications can be applied. The ‘sweat of the brow’ theory sees creative work as entitled to property status equivalent to physical work and informs the common law countries’ approach to copyright. The alternative, civil law approach sees copyright as the rightful reward for an author’s artistic expression of his or her unique personality and perception. For example, Article 2 of the Japanese Copyright Act confers copyright only on works ‘in which thoughts or sentiments are expressed in a creative way’.  

Neither theory really justifies giving monopoly protection to the work of a machine. Without extreme anthropomorphism, unless and until we reach the point where AIs are self-determining in what they choose to do, it is hard to argue that a machine requires monetary reward and legal rights of control over its outputs, in order to produce more work. Nor is there any self-evident reason for giving a form of property right devised for human authors to those corporations which happen to own the processors on which an AI runs. Still less can an AI’s output be characterised as an expression of personality – at least, not until AI achieves the level of HAL 9000, and possibly not even then.  

Nevertheless, Japan’s 2016 Intellectual Property Strategic Program must be right when they concluded that while granting IP protection to all AI-created works (ie information to which the definition of ‘work’ applies) is probably excessive, it is also likely to be necessary, from the perspective of inhibiting free-riders, to protect as some new form of intellectual property those AI-created works which generate above a certain threshold of value. Detailed discussion of what form such IP protection should take needs to occur – preferably at international level, since unless and until embodiment enables AIs to create physical as well as pure information works, AI-created works are most likely to be electronic and therefore exploited online.

Does either theory justify continuing to confer copyright on human-authored works of no greater originality than those produced by the machines? Here, at least, we have a clear answer. The sweat of the brow principle has always been applied without differentiating between great works and those of modest or little value. English courts have accepted that copyright subsists in works as earth-shattering as a betting coupon (Ladbroke Football Ltd v William Hill Football Ltd [1964] 1 WLR 273) or a list of football fixtures (Football League Ltd v Littlewoods Pools Ltd [1959] 1 Ch 637); the courts have never set themselves up as arbiters of artistic worth. And the expression of personality, too, can be seen in works of limited artistic value as much as in masterpieces. Accordingly, creative humans can legitimately expect to continue to be rewarded in copyright. 

But for most now toiling creatively in the commercial, as opposed to the fine art, space, it may not be of much help if, in years to come, the market for many of the relatively anodyne copyright works such as those currently used in packaging, reporting and advertising is substantially taken over by AI-created alternatives. 

Lorna Brazell is a Partner at Osborne Clarke LLP.