Ethical considerations in AI-generated intellectual property

With AI increasing its influence in our day-to-day lives, it’s time to take a fresh look at the man vs machine divide and see how far the lines have blurred. AI is increasingly becoming an integral aide to activities of all kinds, including creative ones, which means that when someone submits a patent application today, there is a fair chance that AI played some role in the creation of the final product. Given that traditional laws only recognize human input as eligible for intellectual property (IP) protection, there are important conversations to be had about the ethics of using AI to generate IP. The goal should be to enable its fair use, but only while keeping human effort, privacy and security rights at paramount.  

The role of AI in IP

Older forms of AI have been used in creative or innovative activities for a while now. These have chiefly taken the form of executing tasks swiftly or processing data rapidly and accurately. With generative AI, however, we have a new class of AI that can simulate human thought and creativity and innovate on its own, in a fashion that can be seen as very close to human invention.  

For instance, some years ago an AI system called DABUS came up with a new type of food container based on fractal geometry. In a landmark decision, South Africa decided to award a patent for the food container to DABUS in July 2021. Similarly, in 2017, AIVA composed an original orchestral piece to open the National Day celebrations in Luxembourg. And there are several AI applications that are already able to generate code on their own, such as Google-backed Bayou or DeepCoder by Microsoft.  

Addressing the IP implications of these AI creations will require significant effort from creators, developers and policymakers to define usage guidelines. Most critically, anything created by AI is only as good as the data used to train it. Thus, it is vital to choose appropriate datasets that do not infringe on any copyright laws and that do not misuse any sensitive or confidential information.  

Understanding India’s IP landscape and the ethical concerns in AI-driven IP

India’s IP laws, at present, are well designed to protect human endeavor but do not contain any provisions to recognise software and algorithms that contribute to IP creation. The Copyright Act of 1957 states that original works expressed in tangible form (such as literary, musical or artistic works) are eligible for copyright protection. The Act also specifies that in the case of computer-generated works, the person who caused the work to be created is the author. On the other hand, the Patent Act of 1970 and the Design Act of 2000 do not recognise a programmer/developer as the owner/inventor in case any innovation was the result of using software or algorithms.  

AI-driven IP presents several problems. Under current IPR laws, only entities with legal personhood (natural or corporate) can be given IP rights. Moreover, there must be clear ownership of the IP for rights to be granted. AI algorithms in their present state of development cannot be described as legal persons, and neither are there any provisions in Indian IP law specific whether ownership can be given to software or algorithms. This also makes it harder to draw clear lines between human and AI creations, how long the IP will be protected, who will be the beneficiary of any remuneration from licensing the product, etc.  

There is also the matter of the data used to train generative AI systems. Very few products have been created entirely by AI acting autonomously. If an AI system creates a movie script or a musical composition, it does so based on the literary and musical data it has been fed with — in other words, existing literary and musical pieces. In such cases, attributing ownership of the product to the AI system raises many questions about where the data was obtained from and whether consent was obtained from those human creators.  

For instance, Getty Images and a group of artists recently filed a lawsuit against Stability AI for using over 12 million images from the Getty database without permission and without compensating the artists. On the other hand, one could well argue that the AI system isn’t simply copying the source material but is crafting something new based on the AI’s unique learning abilities and insights. This leads to further ambiguity about whether AI-generated material can be classed as “creative” or “innovative”. And finally, using datasets to train AI naturally comes with privacy and security risks that could create major problems if users aren’t familiar with how to safely use AI tools. Samsung, for instance, recently leaked trade secret data into a ChatGPT prompt, which means that the data is now accessible to OpenAI, and potentially being used to generate replies to other users of ChatGPT.

India’s strategic response

The National Strategy for AI by NITI Aayog was introduced in November 2018, as a roadmap for the adoption of AI in five public sectors. The strategy coined the mantra “AI for All”, with an emphasis on the responsible use of AI and democratic access to the advantages that AI presents. According to NITI Aayog, the adoption of AI could boost gross value added for the Indian economy by 2035. To make this happen ethically, they have also initiated discussions on the various considerations, such as the need to assign accountability for AI operations or inherent biases in the data that populates the AI. Moreover, the recent Parliamentary Standing Committee report suggests the introduction of a separate category for the protection of the IP rights of AI-based inventions. It also mentions the need for the Department of Promotion of Industry and Internal Trade to review India’s existing IP legislations to accommodate the impact of AI and AI-generated work.

Global comparisons

Several countries have already invested in new guidelines about patent protection in situations where AI algorithms were used to generate material partly or in full. The most prominent example is the UK, where the Copyright, Designs and Patents Act (CDPA) provides for the copyright protection of works that were generated by a compartment “in circumstances such that there is no human author of the work”. Countries like Ireland and New Zealand have prepared similar legislation inspired by the CDPA. On the other hand, many continue to uphold the fundamental idea that copyrighting something requires it to be original, and originality is by default a human endeavor. In 2019, for instance, the Beijing court stated that human production or creation was a prerequisite for obtaining copyright protection.  

Final words

The generative AI landscape is a nascent one and must be explored with caution. It is vital to upgrade the legal system to incorporate the impact of AI — the pre-condition for this, however, is a reassessment of what defines creativity (traditionally defined as a purely human attribute) and what makes a creator a legal entity. Moreover, the new system must continue to protect the rights of creators, whether human or AI, as well as the right to data privacy and security. This will encourage innovators to continue innovating, with or without the use of AI, and ensure that everyone can benefit from innovations that are safe, ethical and fair in their design.

Manish Sinha

---

Manish Sinha is Founder and CTO of PatSeer.