AI Engineering is a study and practice area that integrates the ideas of systems engineering, software engineering, computer science, and human-centred design to produce AI systems that meet human needs for mission success.

AI systems' sophisticated data processing capabilities come to the rescue. This involve gathering, sanitizing, evaluating, deciphering, and keeping pertinent data. These AI-powered tools are capable of sorting through massive amounts of unstructured data to extract insightful knowledge that is essential for making data-driven decisions.

The Copyright Act, 1957, which is currently the Copyright (Amendment) Act, 2012, governs "copyright" in India. According to Section 14 of the Act, copyright refers to the owner of the original creation's exclusive rights to reproduce, publish, broadcast, and sell, rent, or distribute his creative work. This legal right covers all forms of creative and intellectual works, including books, audio and video recordings, photos, scripts, theatre and film production and recording, sculpting, software, computer programs, codes, videos, and graphic arts. These can be produced offline or online in the era of the internet, but artificial intelligence should not be used to create them.

Reverse engineering and algorithmic openness are two concerns that AI systems, especially those that use machine learning algorithms, may bring up. The fair use theory of copyright law may be used to support reverse engineering for goals such as transparency and interoperability.

Fair Use And Its Four Factors

The Copyright Act's Section 107 offers the legal foundation for judging what constitutes a fair use and lists certain uses that may be allowed, including criticism, commentary, news reporting, teaching, scholarship, and research. According to Section 107, when assessing a fair use question, the following four considerations must be taken into account:

• Factor 1 - Purpose and its use:
  As stated in the fair use Act, commercial applications are typically not preferred above nonprofit educational objectives. Furthermore, the Act expressly identifies a number of uses that are particularly suitable for fair use, including commentary, teaching, research, news reporting, criticism, and scholarship. These are commonplace and significant activities within the institution as well. However, use caution—not all charitable educational applications are "fair." A determination of fair use necessitates the application of all four criteria, not just the goal. But restricting your use to a few of these uses will be crucial to establishing fair use. Furthermore, "transformative" applications are more likely to be accepted as reasonable. Uses that contribute something new, have a distinct purpose or personality, and don't replace the original are known as transformative uses.
   
• Factor 2 - Copyrighted work's Nature:
  This component examines how the used work connects to the goal of copyright, which is to promote artistic expression. Therefore, utilizing a work that is more imaginative or artistic (like a song, movie, or novel) is not as likely to bolster a claim of fair use as employing a work that is factual (like a news piece or technical article). Furthermore, it is less probable that using an unpublished work will be seen as fair.
   
• Factor 3 - The Amount and Substantially the portion used:
  Courts consider both the volume and caliber of copyrighted content utilized when evaluating cases under this aspect. Fair use is more likely to be determined if just a little amount of copyrighted material is used in the usage, as opposed to when a big percentage of the copyrighted work is used. However, under specific situations, certain courts have determined that using a work in its entirety is fair. Furthermore, in some situations, it was decided that it was unfair to utilize even a little portion of a copyrighted work since the choice constituted a crucial component, or the "heart" of the work.
   
• Factor 4 - Effect of the use on the copyrighted work's value or prospective market:
  In this case, the courts consider whether and to what degree the unpermitted use impairs the market for the original work of the copyright owner, both now and in the future. When evaluating this aspect, judges look at whether the use is harming the original work's existing market (by driving away sales of the original, for example) and/or if the use has the potential to create significant harm if it were to expand widely.

Case Laws where Fair Use have been applied and its relation to AI generated content:

Authors Guild v. Google

• It is an important case cited by AI developers regarding fair use of copyrighted training materials
• Second Circuit held Google's copying of books for a searchable database as fair use, based on specific facts of the case
• Google used the books to provide factual information, minimizing market harm for copyright owners
• Unlike Google's case, generative AI training does not provide factual information and can act as market substitutes, harming copyright owners who offer licenses for AI training datasets
• Granting a fair use exception would destroy the market for copyright owners to license their works for AI training datasets

Perfect10 v. Amazon

• It may become a battleground for determining transformative versus non-transformative use
• Perfect10 sued Google for using thumbnail versions of its copyrighted images
• The Ninth Circuit ruled that Google's use was transformative as the thumbnails only served as a "pointer" to the full images
• Google's use was not a mere repackaging of copyrighted works but created a new value by providing information about the copied works
• This case differs from AI-generated art and training AI systems on copyrighted works

AI Engines

• AI Algorithm: Machines can assess data, carry out tasks, and make judgments thanks to AI algorithms, which are instructions. It's a branch of machine learning that instructs machines to become self-sufficient learners. Every task that AI completes is based on a certain algorithm.
• AI Data Training: A collection of data, or inputs, known as AI training data is used to educate AI models how to make precise predictions or judgment calls. An AI training dataset comprising photographs of dogs with the label "dog" will be used, for example, if a model is being trained to recognize images of dogs.
• AI Content Creation: Natural language processing technology is essential for creating AI content. This determines how well it can "understand" and comprehend human language similarly to how we do. It uses machine learning methods to analyse text and learn without the need for special programming in order to do this.

Examples of AI Engines

• DALL-E: Using creative briefs, DALL-E might be utilized to produce original, personalized visuals for advertising campaigns. Without relying on stock photographs or a lot of graphic design effort, a marketing team may input exact descriptions of the product, mood, colour palette, etc. and receive personalized visuals. Stated differently, the purpose of compressing text and picture input data into a latent space is to enable deep learning models to process and interpret it. When dealing with DALL-E, the encoder splits textual descriptions into smaller inputs, which are then transformed into low latent representation.
• GPT: For problems involving natural language processing, GPT is an architecture and training approach based on transformers. There are two steps in the training process. Initially, the unlabelled data is utilized to learn the basic parameters of a neural network model using a language modelling aim. It has been applied to content generation, text data analysis, consumer interactions, and repetitive work automation. Research Advancements: In the fields of machine learning and natural language processing, GPT has sparked more study and creativity.

Copyright Infringement Concerns

• Generative AI tools can be used to infringe on a copyright owner's exclusive rights by producing derivatives. Before entering any copyrighted material into a generative AI tool as part of a prompt, permissions may need to be obtained.
• In India, the Copyright Act of 1957 requires an "author" to be a human or a legal person, thus excluding AI from holding authorship rights.

Legal Framework

• Legislative updates: As AI technology develops, politicians may need to modify copyright laws to handle new issues and maintain a just balance between defending the rights of authors and encouraging innovation.
• Moral considerations: AI systems may inadvertently reinforce preexisting prejudices in training data, raising possible moral dilemmas. Copyright laws may need to take AI-generated content's ethical consequences into account.

Important CPC Clauses Concerning AI and Copyright

• Bringing a Civil Suit (Order IV, Rule 1): If AI-generated material violates a copyright holder's rights, the infringement may be brought through a civil suit under the CPC.
• Injunctions (Rules 1-2 of Order XXXIX): The CPC permits judges to grant injunctions to prevent AI engines from violating copyright in the future.
• Damages and Compensation (Order II, Rule 2): The CPC describes the procedures a plaintiff may follow in order to obtain damages for copyright infringement.
• Jurisdiction (Sections 16–20): Establishes the court with jurisdiction over an AI copyright dispute.
• Expert Testimony (Order XXVI, Rule 10A): The CPC permits expert witnesses to inform the court on technical matters in complicated AI issues.

Conclusion
Some who oppose copyright use contend that doing so might diminish the value of creative works and lessen the motivation for creators. But in a world where material is always expanding, the value of human creatives and curators as tastemakers increases. Some artists have successfully used AI technologies, such as Sougwen Chung and Alexander Reben, and they inspire others to do the same. A fundamental idea that promotes learning, innovation, and creativity is fair usage. Even if laws were to alter, it would be foolish to exclude AI from fair use as it would make the nation less competitive and go against the Constitution's core tenet of scientific advancement.

Written By: Akshi Gupta, BALL.B.(H) - Amity Law School, Noida