Welcome to Degen Lawyer's Newsletter, this week we bring you some fascinating news. Every week we serve up curated analysis, hot takes, and expert commentary on all things emerging tech and law, sprinkled with a healthy dose of wacky and meme-worthy. Enjoy!

In this week’s edition:

WHAT’S HAPPENING AROUND THE WORLD

1. SEC gives the go-ahead to Bitcoin ETF’s

   The SEC approved several BTC ETFs, allowing direct BTC investment on regulated exchanges. This follows a court order in the case of Grayscale Investments v/s SEC and marks a shift from past rejections. This approval only covers BTC and doesn’t guarantee approval for other crypto ETFs, but opens doors for future development in various forms. Given that the potential market for BTC ETFs runs into trillions of dollars, SECs nod has serious ramifications on the future of all crypto ETFs.

2. European Banking Authority brings crypto firms under the umbrella of Anti - Money laundering measures.

   EU’s banking watchdog issued guidance for crypto firms to comply with anti-money laundering and terrorist-financing rules. This aims to standardise AML/CFT practices across the EU. The guidelines cover crypto asset service providers and other financial institutions dealing with crypto. Crypto firms have been under increased scrutiny from regulators in the context of money laundering, and the European Banking Authority’s measure reflects the cautious position

3. Central Bank of Nigeria takes a decision to roll back on crypto ban

   Nigeria’s central bank reversed its crypto ban, allowing banks to handle crypto transactions. The move is due to high crypto adoption despite its ban and so as to bring crypto trading under formal regulation to combat financial crimes. Nigeria has found that resistance is futile. Adopting the global trend to regulate crypto activities instead of banning them can result in the country’s crypto industry seeing significant growth.

THIS WEEKS DEEP READ

Generative AI and ownership of generated works by AI, an IPR analysis

Machines and technology, since time immemorial, have played a significant role in the advancement of human progress, improving living standards, life expectancy, reducing travel time, etc. Creativity, however, has been a uniquely human trait throughout history, with technology being a mere aid. Today, technology, more specifically AI, are creators. Generative AI to be exact.

This has led to debates and concerns regarding the laws and regulations surrounding AI. For your read today, we will be focusing on the IPR issues a work from a generative AI faces.

What is Copyright

Copyright ordinarily refers to the legal right given to the creator or the author of a creative work, so that they can fully exploit the benefits of their works and prevent others from unfairly using their work to the detriment of the creator’s rights.

Copyright protection historically emerged to protect artistic expression by humans, with respect to works generated by humans. But with the technological advancements, Artificial Intelligence has become capable of creating original works without direct human intervention, raising questions as to who is the owner of the copyright in AI generated works.

Generative AI and Copyright in India

Generative AI tools scrap vast amounts of publicly available data from pre-existing sources and use this data to generate outputs. There is no primary content generated by such AI, since its outputs are on the basis of the data inputs used in training the AI model. However, the outputs of generative AI are always unique in the sense that it is not an exact copy of any pre-existing work. The variation may be substantial or trivial, based on user commands and the AI tool itself.

An important starting point relating to generative AI, is with respect to the ownership of generative AI end results.

AI as the copyright holder

As per the practice and procedure of the Indian Copyright Office, only a natural person can be an owner/author of a copyrightable work. The Delhi High Court, in the case of Rupendra Kashyap v/s Kiwan Publishing, has, however, held that an artificial person as well is entitled to claim copyright in a work.

Additionally, in the case of Navigators Logistics v/s Kashif Qureshi, the court rejected a claim of copyright over a list compiled by a computer on grounds, inter alia, lack of human intervention. This reflects the general approach in India to vest copyright ownership in natural persons.

Users as copyright holders

The user of the generative AI might have a case for claiming copyrightability in the works generated by AI, since it is only upon user command that such a work is created. Section 2(d) of the Copyright Act stipulates that the author of an artistic or literary work is the ‘person who causes the work to be created’. As explained in the Joint Committee Report on the Copyright (Second Amendment) Bill 1992, the proviso was specifically inserted through the 1994 amendments in consideration of the rapid developments in technology, including artificial intelligence. It was enacted to reflect an acknowledgement that technology is capable of giving new ideas, apart from what is fed into the system.

But a reference to the terms of services of most, if not all, generative AI platforms indicates that the user of the generative AI does not enjoy ownership of the outputs in an absolute sense. A necessary contractual precondition which a user has to accept before using the AI, is to agree to not commercially exploit the outputs of the AI and restrict its usage for personal purposes. For instance, the famous text based AI generator, perplexity AI, mentions in its term of service that users are granted access for their personal and non-commercial use only.

Generative AI and IPR rights of people whose works have been used as training inputs

Authors often challenge generative AI platforms because their works were used to train the AI without their prior permission. In a suit filed by illustrators in the Northern District of California titled Andersen v/s Stability[, the artists have sued the platform for using their original works without a license to train the AI in their styles, which allows the AI to generate works that are not sufficiently transformative from the existing works protected by IP. Restating the claim in the Indian context, the works generated do not have sufficient originality to be calleunauthorisedd derivative. Still, they would instead be treated as an unauthorise adaptation of protected works, resulting in infringement of such protected works.

Comedian and author Sarah Silverman are suing OpenAI and Meta over copyright infringment, alleging that OpenAI’s ChatGPT and Meta’s LLM were trained on illegally obtained datasets containing their works, from ‘shadow library’ websites such as z-library, Bibliotik, etc.. This opens up a new dimension in IPR issues, which not only talks about unauthorised use of copyrighted work, but obtaining such copyrighted work from unauthorised sources int eh first place.

Even Getty, a famous image licensing service has filed a lawsuit against the creators of generative AI at stable diffusion, alleging that they have improperly used their photos, violating the trademark and copyright it has in its watermarked collection of photographs.

While these cases are undergoing trial in foreign judiciaries, the outcome is expected to hinge on the interpretation of the fair use doctrine, specifically whether there is a transformative use of copyrighter material in a manner for which it was not intended. A clarity regarding whether such works can be considered derivative works is also expected.

Conclusion

There is no one size fits all solution to the question of ownership. Legislative guidance is required on this point, as all parties involved in the creation of a work by generative AI might have an equally competitive claim of authorship over such content. Each party’s claim finds validation and rejection in the same breath, based on different legal principles. The ownership can be decided using a utilitarian approach, exploring which of the parties have played the most significant role in the generation of the AI generated work.

DID YOU KNOW?

We all have heard the term ‘Language models’ in the context of generative AI. But did you know that the use of language models goes back before AI was even conceptualised? Early language models were basic statistical models that used probabilities to predict the likelihood of a given word based on the words that came before it. For example, the n-gram model, which was referenced as early as 1948 by mathematician Claude Shannon, uses probability and statistics to determine the likelihood of a word by those that come before it.

If you would like to receive regular updates on all things regulatory and policy in the emerging technology space, please click the subscribe button below and follow us on our socials and share it with others you think will enjoy it!