Artificial Intelligence in Indian Arbitration: A Framework for Ethical AI Integration
INTRODUCTION:
Artificial Intelligence (AI) can learn like humans. Growing from every new set of information and using it in the very next problem. With the help of machine learning algorithms, AI systems possess what other computing machines could never have, the power to learn. Not just from its present command, but from every command ever input into it on any of its portals. Putting it in the mechanisms of dispute redressal, it is as if using a Ferrari to mow the lawn. Too much for too little? Not exactly. AI in all its strength and knowledge may still not be enough to tackle the complexities of the Indian dispute resolution procedures, especially while focusing on Alternative Dispute Resolution (ADR) mechanisms such as Arbitration.
India has, with its due diligence, positioned itself to become a functional International Arbitration hub through its Draft Arbitration and Conciliation (Amendment) Bill, 2024 (hereinafter referred to as “the bill”), along with multiple technological initiatives that focus on maintaining strict judicial integrity while also paving the way for digital innovation and inclusivity. But, does integration of AI systems in the ADR justice framework seem practicable? These questions remain unresolved. But still, this step of introducing the bill can be considered innovative. This piece will delve further into the issues faced by Indian AI governance and what will hinder India’s path to global success.
STATUTORY STEPS AND THE DISTANT GOAL
Removal of human agency in a dispute resolution process, such as arbitration, is indeed a questionable step. The experience of arbitrators is one of the primary preconditions on which the process as a whole is trusted, along with being quick and also pocket-friendly. AI is set to overtake tasks pertaining to the administrative work of ADR agencies, along with scheduling or notifications of sessions. Even still, the bill does not mandate the use of explicit “Artificial Intelligence” in its proceedings, but inputs pertaining to the usage of means such as audio-video measures as U/s 42(d) displays an intention of probable inclusion of digital innovation. The government of India is taking drastic measures for the digital growth and innovation in the country with the help of AI. The India AI Mission has been entrusted with a budgetary allocation of above ten thousand crores, aiming towards the mitigation of bias in AI algorithmic systems and the ethical certification mechanisms of both inputs and results.
Even more explicitly, NITI Aayog has been advocating for its ground-breaking report titled “The ODR Plan for India,”wherein AI-powered online Dispute Redressal mechanisms are proposed to take over. Here, as is proposed, the AI algorithm will act as a fourth party to the dispute to counter any arbitration challenges. NITI Aayog itself agrees that implementation of this in India will lead to similar results as have been showcased in other countries who have already been using this, such as the UAE. Its success is primarily driven by its cost-effective and convenient nature. But there is a catch. What is actually a primary concern is privacy and confidentiality. Technological reassurances and prior approvals for technology usage might help in creating a confidential and comfortable environment, but how far can technology be trusted?
DIFFERENTIAL ANALYSIS AND THE LEGAL DISPARITY
The first question of trust is Algorithmic Bias. As explained above, AI keeps on growing from its past commands. Such commands may, in cases, pertain to ingrained biases such as caste, race, socioeconomic status, etc. If AI is as such trained on its flawed nature of data itself, it will lead to the perpetuation of amplified existing social prejudices, leading to highly discriminative outcomes. Violating Article 14 of the Constitution, which promises equality before the law, is surely a framework that needs to be evaluated. Section 31(3) of the Arbitration Act brings a requirement as to the explanation and reasoning behind an arbitral award, something which AI will never be able to encompass in a human-like comprehensive level while determining highly important issues. This will directly hamper the right to appeal and set aside an arbitral award under section 34, as without understanding the logic and method the algorithm chose to pronounce the award, the parties will not be able to challenge the decision.
The above issue can be tackled. Deep learning systems are coded to process information through hundreds and thousands of data nodes, which makes it impossible to track the precise source and path from the input to the output. Re-coding the evaluation process may be a way out. However, most AI developers have proprietary algorithmic secrecy, which allows them to refuse to disclose the AI’s internal workings due to reasons of competition. India thus, has a massive regulatory gap which needs to be bridged. India even today lacks AI explainability standards, which regulate requirements for the explanation of algorithms in the legal context. Furthermore, as a tag along to this, the Arbitration Council of India (ACI) remains silent on the development of AI transparency guidelines, and even this amendment bill contains no explainability mandates for AI systems.
Globally, in countries where the implementation of AI in ADR mechanisms has in fact, shown acceptance and functioning, there remain these guidelines which India lacks. By recognizing the digital processes, there is acceptance of technological prowess, but there is still no mention of AI governance provisions. As of today, even the Supreme Court is silent on AI governance mechanisms in arbitration. For example, section 34(2A) of the Act mentions the setting aside of awards and patent illegality stands as one of the reasons, which again brings into question whether AI-generated reasonings can be considered patently illegal under this ambit.
In Singapore, different from India, it is arbitration’s “gold standard”. This is so because there exists a formal system of AI-integrated features in filing, document storage, and real-time case tracking, along with the power of the registrar to direct all communications through digital means from the commencement of arbitration. The SIAC has provided its latest 2025 rules regarding AI integration in arbitration, which can in fact be used as a standard for measuring India’s current position and possibility of growth. In a different fashion, the UK’s recent judgement in the case of R (Ayinde) v. London Borough of Haringey & Al-Haroun pointed out that freely available generative AI models, which are trained on a large language model, such as ChatGPT, are not capable of conducting reliable legal research through Justice Jeremy Johnson’s comments. It is notable that this strikes a threat to possible Indian AI governance models but it also gives a space to learn.
WAY FORWARD: TWO STEPS FORWARD, NO STEPS BACK
At the end of the day, to trust AI in the proposed ODR mechanisms can only occur through specific allegiances. The major problems encountered by India are transparency and bias. AI cannot be trusted to be transparent in its reasoning and understandings without the inclusion of human agency.[E12] In this same issue, the EU’s GDPR under Article 22establishes that the data subject, i.e., the party, shall not be subject to any decision made wholly and solely by an automated processor producing binding legal impacts. It was even furthered in the case of SCHUFA Holding and Others v. Deutscher Sparkassen- und Giroverband, where article 22(1) operated as an automatic prohibition and restriction. The burden shifts to the owner of the AI system instead of the individuals to invoke the right. But what is most fascinating is the ECJ’s stance when it mentioned that decisions produced by AI systems with significant legal impacts, involving human agency becomes rather mandatory. Logic is to precede every decision. The CIArb guidelines of 2025 also call for mandatory transparency and encourage parties’ consultation before the use of AI tools.
The success of any AI governance evolution in India will pivot on addressing AI transparency challenges. India’s commitment to becoming a global arbitration hub by 2030 will be possible only through focus on institutional arbitration. As such, for the fulfillment of this dream, India is bound to implement a comprehensive and multi-layered framework addressing all dimensions of transparency. Human oversight protocols, as is there in the EU, will give a boost to India’s AI governance models. Not only this, but the comprehensive DPDP Act of 2023’s compliance shall also be necessary with features such as the collection of minimal and only necessary data along with explicit consent mechanisms for data processing and inputs. Lastly, to mitigate bias and promote arbitrational fairness, standards such as disparity thresholds shall mandate measures to prevent algorithmic bias and AI-based discrimination. This is the opportunity for India to demonstrate to the world its technological innovation in fair justice. The stakes are high: done right, AI can democratize justice and position India as a global arbitration hub. Done wrong, it risks undermining fairness, security, and trust in India’s entire dispute resolution ecosystem.
*Aadit is a third year student from Rajiv Gandhi National University of Law, Patiala.