AI tools in Indian courtsCEFR B2
5 Dec 2025
Adapted from Sakkcham Singh Parmaar, Global Voices • CC BY 3.0
Photo by Rishu Bhosale, Unsplash
Indian courts are adopting a range of digital technologies as they confront a backlog of several tens of millions of cases. The government and the Supreme Court are pushing modernization through Phase III of the e-Courts project and by investing in future technologies such as AI and blockchain. Digital change began with the e-Courts programme in 2007, which introduced e-filing, digital cause lists and online judgments; Phase III now targets machine learning and language technologies for digitised judicial information.
SUPACE, the Supreme Court Portal for Assistance in Courts Efficiency, is a major innovation. It assists judges and research staff by identifying facts, suggesting precedents and drafting outlines, but it does not decide cases. Language tools include SUVAS, developed to translate judgments from English into other Indian languages, and pilots that convert local-language judgments into English. Automated transcription, first used in constitutional matters, has produced near-real-time, searchable text since 2023.
In 2025 the High Court of Kerala ordered subordinate courts to use the AI-enabled speech-to-text tool Adalat.AI to record witness depositions from November 1, 2025. Adalat.AI was developed by a start-up with research links to Harvard and MIT; the order allows judges to use alternative platforms vetted by the High Court’s IT Directorate if the system fails. Supporters say AI can reduce human error in transcription, catch basic e-filing mistakes, shorten hearings and help prioritise urgent cases, potentially improving access to justice in remote districts.
But judges and scholars have warned of real risks. The Delhi High Court in 2023 refused to consider arguments that relied on ChatGPT, and found filings that contained non-existent cases and misquotes. Concerns include model opacity, bias in training data that may reflect caste, gender, class or religion, and the danger that AI will amplify discrimination. Judicial records contain sensitive personal data, so guidelines discourage uploading such material to public cloud tools; the Digital Personal Data Protection Act, 2023 applies to automated processing. Court policies seek a middle path: Kerala’s AI policy treats AI as an administrative tool, bans generative AI from drafting judgments or predicting outcomes, and forbids external platforms that require uploading confidential information. The Supreme Court has set up an AI Committee and is working with institutions such as IIT Madras. Experts recommend audits for bias, mandatory disclosures when AI is used, better infrastructure, judicial training and routes for litigants to challenge AI’s role, while judges aim to keep humans in charge and treat AI as an assistant rather than an oracle.
Difficult words
- backlog — unfinished legal cases waiting to be heard
- e-filing — sending legal documents to a court electronically
- transcription — written record of spoken words or speech
- precedents — earlier court decision used as a legal example
- depositions — spoken testimony given under oath for record
- bias — unfair preference or prejudice in data or decisions
- opacity — lack of clarity about how something works
- vetted — to check and approve something as acceptable
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- How should courts balance the potential benefits of AI (like faster transcriptions) with the risks (such as bias and opacity)? Give reasons.
- What steps could improve access to justice in remote districts using digital tools, and what practical problems might appear?
- What safeguards would you suggest to protect sensitive personal data when courts use AI and cloud services?