Continued from Part 1 –The AI Conundrum: Copyright, Creativity and Challenges in the Age of Generative AI
Another major copyright concern that arises from the development of works by GAI is regarding the “training”/ “feeding of information” to the GAI by the developers. GAI such as Dall-E are capable of creating artworks that emulate the style of various artists and generate a novel work[1]. The reason that GAIs are adept at emulating the style of earlier artists/authors, is because they have been provided access to the data pertaining to such artists/authors combined with the machine learning process, the GAI provides a response that may be suitable for its users. Obviously, several GAIs are being used by their developers for commercial purposes thereby resulting in them earning considerably through their programs. On the other hand, artists/content creators are exasperated by this development since, their original works are being used as training materials for the GAI, without their consent and of course, without paying any license fees. Therefore, in January 2023, in the U.S.A., artists Sarah Andersen, Kelly McKernan and Kayla Ortiz filed a class action suit for copyright infringement against Stability AI[2]. They claimed that their works were used to train the AI without their permission thereby infringing their copyright. Furthermore, it was claimed that the works generated by the Stability AI’s software were derivative works infringing the original creator’s copyrights. The defendants in this suit have contended that the program was trained with 5 billion compressed images and the resulting work developed by the AI was unique and not infringing any copyrights. In the hearing of this matter in July 2023, the judge has seemingly indicated that he is inclined to dismiss the plaintiffs’ complaint and will give them an opportunity to file a new complaint providing more clear facts/allegations against the GAI developing companies[3]. This suit is one of many more that have since been filed in the USA, by various authors/content creators claiming infringement of their copyrights. The authors’/content creators’ aggravation is quite justified however, it is also important to consider whether this unauthorised use of their works may fall in the category of fair dealing.
In or around 2005, a somewhat similar concern had arisen in the case of Authors Guild, Inc. v. Google, Inc.[4] (“Google case”) wherein Google had started its digitization process by scanning and identifying online texts without any license from the authors/publishers of the books that were digitized. Google contended that it would only make certain portions of the work accessible to public (And not the work in its entirety). In the suit, it was finally decided that Google’s digitization was transformative in nature and was well within the ambit of fair use doctrine.
More recently, in the case of Google LLC v Oracle America, Inc[5], (“Google-Oracle case”) Oracle claimed that Google had used 11,000 lines of code from the Java SE application, owned by it, for developing its Android operating system. It is pertinent to note that JAVA is a language which in 2005 was familiar to programmers and the use of this language to develop implementation code for Android systems (by Google) played a major role in the growth and advancement of the Android OS, globally. The U.S. Supreme Court case ruled that Google’s use of the codes from Java SE application amounted to fair use. This decision was lauded by programmers since it promotes developing of interoperable operating systems and supports software innovation.
In India, one of the most discussed copyright suits of recent times University of Oxford v. Rameshwari Photocopy Service[6] d (“DU photocopy case”) decided by the Hon’ble Delhi High Court. In this case a few publishing houses filed a suit for copyright infringement against the defendant, alleging that by photocopying parts of their books they were infringing the copyrights of the owners. The defendants contended that only a few chapters/sections of the books were photocopied to form. Part of the reading material for the students (from the instructors). The Division Bench of the Hon’ble Delhi High Court ruled that copyrights granted under the Act are also subject to exceptions under the law. In this case, whilst photocopying indeed amounted to reproduction of an original work, it was pertinent to note that the said act was being done for the purpose of imparting education which was a recognised exception under Section 52 of the Act and therefore, there had been no infringement of copyright.
The common finding in the Google case , Google-Oracle case and the DU photocopy case is that certain acts of copyright infringement when serving an educational or public information related purpose or for fostering innovation, may fall within the ambit of copyright exceptions namely -fair use (USA) or fair dealing (India). In my limited view, the principles laid down by the aforementioned cases may also be applied to the issue of information/data protected under the law of copyrights, that are being relied upon for training/developing a Generative AI program. The rationale behind this is that the copyrighted works are being used to foster innovation and technical advancement in society, which is the pinnacle of IP laws. Furthermore, as argued by representatives of Stability AI, the results shared by the GAI do not plagiarize or replicate identical work used to train it; in fact, it is a novel/original work in itself.
What would a lawyer’s work entail, if things were this simple so, of course, comes in a more concerning and complex issue that has arisen with the advancement of GAIs is when the programs are advanced enough to simulate the voice/likeness of renowned artists. For instance, earlier this year a user developed an AI that developed a song simulating the voice of famous artists such as Drake and The Weeknd, this song was released on Spotify and Apple Music and was successful in raking up millions of streams, before being taken down by the Universal Music Group, the label managing the 2 artists[7]. Ever since, social media platforms are flooded with GAI developed songs, imitating the voice of famous artists of the past (Freddie Mercury, Michael Jackson etc.) This new development, is indeed very chilling because most members of public are unable to distinguish an AI generated voice from that of the original artist. Moreover, these AI generated voices, are capable of replacing the original artists and are directly competing with the artists in the same industry. This exploitation of an artist’s voice (via the medium of a GAI) amounts to exploitation of the artist’s personal rights, also known as publicity rights or personality rights.
In India, personality rights stem from the right to privacy, a fundamental right, granted under Article 21 of the Indian Constitution. Very recently, in the case of Amitabh Bachchan v. Rajat Nagi and Ors.[8], the legendary movie star filed a suit for the protection of his publicity rights against scrupulous entities running a Kaun Banega Crorepati (KBC) fraud by using his name, voice or other attributes exclusive to him. Furthermore, it was also alleged that there were various domain names being registered using his name/surname and other names attributed to him such as “AB”. The Hon’ble Delhi High Court found merit in the complaint of the megastar and accordingly passed an order restraining defendants from infringing the personality rights of the megastar. Applying the ruling of this case, to the issue of GAIs imitating the voice/likeness of celebrities, it can be safely argued that such utilization of GAI will definitely attract multiple law suits from the artists whose likeness is being used without their authorisation.
Whilst technological advancements have made our lives easier than ever, but just like all good things have a price to pay, it appears that the challenges posed by AI/GAI will have to be addressed sooner by all the state governments to ensure that the programs that were intended to assist us, end up causing more complications than ever. Putting it in a more pop culture context, right now the AIs are quite similar WALL-E (directed by Andrew Stanton) but if we don’t regulate the laws, update and accommodate ourselves to this new reality, our future with AI could be similar to The Terminator (1984) directed by James Cameron, ok so that may be a bit of a stretch…. but may be Her (Samantha -the AI) directed by Spike Jonze, who knows?
Endnotes:
[1] Goldman, S. (2022) Who owns DALL-E images? Legal AI experts weigh in, Venture Beat. Available at: https://venturebeat.com/ai/who-owns-dall-e-images-legal-ai-experts-weigh-in/ (Accessed: 02 August 2023).
[2] Andersen v. Stability AI Ltd, U.S. District Court for the Northern District of California, №3:23-cv-00201.
[3] Brittain, B. (2023) US judge finds flaws in artists’ lawsuit against AI companies, Reuters. Available at: https://www.reuters.com/legal/litigation/us-judge-finds-flaws-artists-lawsuit-against-ai-companies-2023-07-19/ (Accessed: 02 August 2023).
[4] 721 F3d 132 (2nd Circuit 2015)
[5] 141 S.Ct. 1183
[6] 2016 SCC Online Del 6229
[7] AI-generated song imitating Drake & The Weeknd removed from streaming services (2023) Economic Times. Available at: https://economictimes.indiatimes.com/magazines/panache/ai-generated-song-imitating-drake-the-weeknd-removed-from-streaming-services/articleshow/99599387.cms?from=mdr (Accessed: 03 August 2023).
[8] 2022 SCC OnLine Del 4110