Mint & Co Industry Insights: AI and Computer-Generated AI
Nicola Hartley
Founder & CEO Mint & Co - Legal & Business Affairs for the creative industries. Founder - Mindful Mint (wellbeing & personal development programmes). Co-Founder Path to Purpose Retreats. Poet, artist & wellbeing warrior
Turbulent times for intellectual property policy, regulation and legislation
By Jo Rosenfelder , Mint & Co Legal & Business Affairs Consultant
“And when everyone’s super, no-one will be” threatens Syndrome – the villain of “The Incredibles”, as he unveils his new computer program that can create original and brilliant art, music, literature.? Whilst this might have been a distant nightmare when The Incredibles was first released twenty years ago, today it is a reality, with the products of Artificial Intelligence programs being challenged in the courts by the world’s most powerful owners of creative rights.
1.???? Introduction
It is almost impossible to overstate the impact of technology and Artificial Intelligence (“AI”) on our lives over the past 20 years – whether in the home, workplace or at leisure. New effects of this revolution are still emerging, with broad and deep implications on our society.
Whilst governments and policy makers grapple with the issues surrounding this technological tsunami, regulators and lawyers are tasked with finding legal solutions in an era of immense change. Legal frameworks, definitions and precedents in case law that have governed the treatment of intellectual property rights for decades are coming under wholesale challenge.? Existing statutes and precedents are simply unsuited to the new landscape.???
And beneath the political and legislative conundrum, there are also significant moral, ethical and legal concerns, as outputs of Computer Generated AI (“CGAI”) are becoming increasingly sophisticated. ?Governments, social scientists and ethicists are being asked to decide who and what rights should and shouldn’t be protected in the age of CGAI. ??There are issues across several fields, including:
These are just a few of the underlying issues have been the focus of extensive government thinking for several years already, as policy makers struggle to create new frameworks to address technology developments which are still moving at a dizzyingly fast pace.? But as they struggle to pin down new paradigms in a dynamic new landscape; inventors and creators continue to develop and publish work, and computer scientists roll out increasingly sophisticated CGAI programs.? New works are necessarily being created and published within the existing legal frameworks, which were not designed for this brave new world.? How can we advise in these fast changing times?
2.???? What is Artificial Intelligence (“AI”)??
Although, to many of us, the idea of AI seems a very recent invention, the concept of AI was first coined nearly 70 years ago for a Dartmouth Summer Research Project convened by two professors, John McCarthy and Marvin Minsky. They defined AI as “any task performed by a program or machine that, if a human carried out the same activity, we would say that the human had to apply intelligence to accomplish that task”.
Whilst AI has not necessarily developed in the way that McCarthy and Minsky predicted (or necessarily wished), their definition continues to be generally useful today.? The term AI is broadly applied to systems that typically demonstrate behaviours associated with human intelligence, such as planning, learning, reasoning and problem solving.? Although we’re not quite there yet, scientists predict that it won’t be long until AI will also be able to develop some representation of social intelligence and creativity.
AI in use today – “Weak” or “Narrow AI”.
Almost all AI in general use today is still a version of “Narrow AI” or “Weak AI”– that is, AI based on machine learning.? “Machine learning” uses statistical methods to process huge numbers of data sets to arrive at decisions, which often look like human decisions. ?
In contrast to “Weak AI”, “Strong AI” is the AI often characterised by the sorts of AI we see in sci-fi movies – of robots acting and behaving like humans flexible and capable of carrying tasks that require reason and creativity, with ‘personalities’ based on a variety of accumulated experience, essentially founded on “neural networks” which work in a similar way to the human brain.??
As Weak AI becomes more sophisticated, commonly used narrow AI programs are looking more and more like they’re Strong AI, with outputs that seemingly extend the limited tasks of narrow AI. ?However, and although seemingly much more sophisticated than their initial versions, widely used AI programs including Chat GPT, Bard, Midjourney, DALL.3 E and Copilot are still, essentially, Weak AI programs. ?Their increased power is based on ever larger data sets, which has similarly increased the sophistication of their responses. They use hundreds of millions of data sets to predict responses, and deliver an optimised output, which is based on their statistical analysis of a huge number of responses.? But they are still only machine-learned responses, leading to anomalies if and when less commonly switched variables are inputted.? It is the application of this sophisticated “weak AI” that most lawyers are currently having to address.
3.???? Intellectual Property law and use of CGAI
Essentially, intellectual property arises in AI programs in three distinct areas:
As media lawyers (looking primarily on broadcast media), the focus of the AI challenges that we meet are mostly on AI outputs, (although IP lawyers are also looking and grappling with the IP protections both for Core AI and the AI method).? ?Even though prompts can be very simple, might the AI inputs (the ‘prompts’) be deemed to be sufficiently original to attract copyright protection in the AI Output?? Who owns the IP rights in the outputs of an AI program?? And what are the issues that might challenge its use??
In the UK, whilst parliament works to find new regulation and policy, creators and IP owners are still applying the existing statutes to decisions over copyright ownership – primarily the Copyright, Designs and Patent Act (“the CDPA”) 1988 and the Copyright and Rights in Database Regulations 1997.
领英推荐
Originality and authorship?
Copyright protection under the CDPA is based on the application of two concepts –
So, who is the author of the output of an AI program, and is the output “original”?
In the UK, and in contrast to other jurisdictions, it is generally suggested that the test for copyright ownership, is quite low – copyright will subsist for an author, provided
The rights defined in the Copyright and Rights in Database Regulations 1997 additionally protect the owner of a database against the extraction or re-utilisation of ‘the whole or a substantial part of’ the contents of the database, provided there has been a substantial investment in ‘the obtaining, verification or presentation’ of the contents of the database.
Computer generated IP and the CDPA and Database Regulations
In most jurisdictions “computed generated works”, are not covered by copyright.? However, and showing more foresight than other jurisdictions, from the beginning the CDPA has explicitly recognises “computer-generated” works as work “generated by computer in circumstances such that there is no human author of the work”.(s178)
Section 9(3) of the CDPA ?provides that the author of a computer-generated work is the person “by whom the arrangements necessary for the creation of the work are undertaken”. This highlights some of the uncertainties around existing legislation – as, using that definition, the “author” could be any one of a number of parties including:
Issues of authorship are a fundamental question that surround CGAI.? Courts have suggested that it might break down to the level of contribution of each party (creator of the software or the database, teacher of the software, or the user through prompts).???
Recent years have generated some interesting cases around copyright and patent ownership, authorship and use of AI both in the UK and the US.? I’ve included a few examples here, just for illustration:
Naruto v Salter No.16-15469 (9th Cir, 2018)- in this US case, PETA (People for the Ethical Treatment of Animals), claimed copyright on behalf of a monkey (called Naruto) to a selfie taken by the monkey himself, with a camera belonging to Slater, a wildlife photographer.? The court ruled that non-human generated content inherently in the public domain, - and whilst it questioned whether animals can hold legal rights, it is also includes one of the earliest reviews around ownership of AI generated content..
Stephen L Thaler, v Comptroller-General of Patents, Designs and Trade Marks [2021] EWCA Civ 1374:? In this case Dr Stephen Thaler created a picture using his AI engine , and applied for a trademark, listing the AI engine as the author of the work, stating that the work was created by the AI machine.? The Court of Appeal ruled that a machine is not an inventor for the purposes of the UK Patents Act 1977.? ???The case was also being argued by Thaler as a copyright case in the US – where it has so far been refused as the US copyright office requires “human authorship.” It is now being argued before a federal judge.
Getty vs Stability Diffusion –? This case, which is currently going through the UK courts, involves a challenge by Getty against Stability, which “scraped” the internet for millions of images to help train its machine-learned photographs.? Getty is still suing for use of its copyright.? The Getty case is testing several aspects of AI and copyright – particularly including questions over jurisdiction, fair dealing (Stability is arguing an exception for “pastiche”).Essentially, this case is evolving into an early testing ground for the limits and power of copyright holders in the brave new world of CGAI.
Another highly publicised case going through the courts at the moment involves three global media companies - Sony, Warner Brothers and Universal, who, together, hold the rights to many of the world’s leading music catalogues.? Each of them is challenging two newly launched AI programs, Suno and Udio.? Both Suno and Udio have simple music programs that, in response to the simplest of prompts, will produce songs similar to that produced by the most iconic bands of recent years.?
Like Stability, Suno and Udio are machine learning programs, which essentially ‘scrape” the internet by learning from the digitised character of the works stored by the catalogue owners online.?
Together, the cases filed by Getty against Stability and the music titans against Suno and Udio highlight some of the copyright issues challenged by AI –presaging seismic changes across almost every area of our lives.
4.???? How to protect or manage the use of CGAI in IP contracts?
As highlighted in recent cases in the UK, US and Europe, essentially, in all jurisdictions, existing statute and case law protections for copyright, patents, design rights and database rights do not apply clearly enough to CGAI.? To support their clients, during this period of intense change both in technology and regulation, lawyers and advisers need to find ways to protect the owners and creators of these new works.
Current IP law was not designed for AI and is obviously not suitable for it. So, whilst governments decide what to do, how should lawyers and advisers approach issues around CGAI??
Without a clear regulatory framework, lawyers and contract advisers ?should instead use contractual agreements including licensing and collaboration agreements to ensure that express contract terms are used to transfer ownership or grant licences to use any AI including CGAI and any related IP that may be required.
It is strongly advised to ensure that licences are sufficiently widely drawn to ensure that any CGAI that may be included in any agreement includes warranties and indemnities to cover for the use and any possible infringements of any CGAI by the other party (whether in the Core AI, the AI Method or the AI Output).
Several AI companies are already acknowledging the issues, and most of the leading AI programs including those of Microsoft, Google, Amazon and Open AI are offering some ?protection for customers against copyright infringement lawsuits, particularly for subscribers to paid-for programs.? ??But, whilst the provision of these clauses does ostensibly appear to provide some comfort, these protections may be quite limited, and often include significant exclusions and may be subject to a very low cap. ?As ever, it is essential to read the small print!
The intensity of this period of technology development and the huge issues thrown up by AI mean that any clear legislative framework is unlikely to emerge for some time.? In the interim, those involved in the negotiation and drafting of any contracts where CGAI is being transferred or licenced in any way, should ensure that they appreciate the underlying risks and infringements inherent in CGAI and ensure that they include relevant contractual terms, including appropriate warranties and suitably drawn indemnities (paying particular regard to exclusions and caps), to protect their clients’ interests.