3 minute read

Copyright for AI works still unresolved

Werina Griffiths

Adams & Adams

If a userwere to provide an artificialintelligence (AI)program, suchas ChatGPT, with a storyline and it produced a bestselling novel, would that novel be protected by copyright? And ifit is, whowould own those rights?

There has beenan explosion of interest in AI platforms following the launchof OpenAI’sChatGPT servicewhich attracted more than 1-million usersinthe firstfourdays after its launch.

Headlines that ChatGPT canmake giftrecommendations,debug code,passan exam, as well as write essays, academic articles,comedy routines, recipes and even music,notonly holdtruebut demand consideration of whatishappening, howitis happening andwhat the potentialimplications arefor IP rights and their owners.

ChatGPT is a chatbot technology, whichmeans itis a computer program that uses AIand naturallanguageprocessing to understand questions and then automate responsesto them.Chatbots are powered by large amounts of data and machine learning techniquesthat enable themto makepredictionsregarding themost accurate answer to the question, which is then formulated and provided to the user.

While the first of its magnitude, ChatGPTis notthe only AIplatform capableof generating content in this way. Prior tothe launch of ChatGPT, OpenAI had already launchedan AIgraphicstool called DALL-E which can human mind are protected by thedifferentforms ofIPand, in particular, copyright.

Putting the theories to the test, the artwork accompanying this article (let’s call it “The Robot”) was created using OpenAI’s DALL-Eprogram by entering the text prompt “ a robot painting on canvas, expressionism”. Within what felt like seconds, the hype around thetechnology became very real when the resultof thetextprompt delivered something that is, at leastin principle,certainly copyright protectable.

ARTIFICIAL ART?

The Robot was created using OpenAI s DALL-E program by entering the text prompt a robot painting on canvas, expressionism convert textinto graphics. One can only imagine that the possibilities for creators become endlessif allthat is required is a simple text prompt.

Traditionally, workscreatedthrough theintellectual or creativeefforts ofthe

Oneof thegreatestbenefitsof copyrightprotectionis that itcomes intoexistence automatically. Themoment a workis createdthatqualifies for copyright protection, it is protected. Therefore,if AI generated contentsatisfies the requirements for copyrighttovest, it,too,couldbe protected in whichcase the rights in and to that work will belong to someone.

The Copyright Act in SA differentiates between the traditional authorialworks (eg literary, musical and artistic) and such works that are computer-generated . The Robot would qualify as a computer-generated artistic work. So far so good.

The next step relevant to this discussion isto determine whether thework was “original” in its making and, to answer thatquestion, we havetoconsider totheskill, effort and labour expended by the author in the creation of thework. AIgeneratedcontent has no human author, but the Copyright Act dictates that the “author” of acomputergenerated work is the person responsiblefor makingthe arrangements for the creation of the work.

The meaning of “making the arrangementsnecessary for the creationof the work” isnot entirelyclear andit is likely here that the debate will arise.

The firstpotential argumentisthat thedeveloperof the AI could bethe author, as defined, forhaving made arrangementsthat werenecessary for the work to be created.

However, when it comes to programs suchas ChatGPT,DALL-E andothers whichgenerate worksby making independent decisions in determining what the work should look like, the results generated by the programs arenot fixed,nor are they designed by the developer.In fact,the resultscannoteven bepredictedor expectedby thedeveloper anddepend inthefirst instanceon somethingthatis conceptualised by the user.

The proximityof the developer to the work createdascompared totheinput of the user leadsto a second potentialargument, whichis that the user could be considered the person responsible for making the necessary arrangements for thework to be created.

Making Of The Work

Once theauthor hasbeen identified, it isnecessary to consider whether the author’s efforts in creating the work were sufficient to renderthe workoriginal inits making, suchthat itwill be protectedby copyright.This will involvea factualinquiry around themaking ofthe work in each instance.

The answers to the question of copyright ownership over AI generated content are not straightforward and require careful consideration. While thetechnology gains speed and attention,and legal mindsexamine itsimplications, the take-outwould be toembrace thepossibilities with the awareness that they are not risk free.

As forThe Robot, OpenAI’s terms currently state thatany rightsarising fromthe creationofworks areassigned tothe usersubjectto compliancewiththe terms ofthe useragreement.

The terms also contain a licence in favour of OpenAI to continue usingthe content generated and a warning that the AI programmay generate the same or similar output for other usersgiven thenature of machine learning.

TheRobot thereforefinds itselfin apositionthat isneither certain nor unencumberedwhichpoints tothefact theuse ofAIto generatecontentfor commercialexploitationshould beapproached with extreme caution.

This article is from: