India | Increasing use of AI across the justice system | Oxford Institute of Technology and Justice

Tools

Tasks

Users

Law enforcement | Prosecutors | Courts | Defence | Victims

Scope

Nationwide

Training

Yes, for judges and law enforcement staff

Regulation

No dedicated legislation targeting the use of AI in courts. Local policies and guidelines, including from the Kerala High Court and the Bombay Bar Association, are starting to emerge to restrict the use of AI tools in court. General rules of evidence and of criminal procedure and data protection laws may also apply

Cases

The Delhi High Court rejected a ChatGPT response relied upon by the plaintiff in Christian Louboutin SAS v The Shoe Boutique, CS(COMM) 583 of 2023, a trademark case, confirming that responses from LLM based chatbots such as ChatGPT cannot form the basis of deciding factual or legal issues. The Court observed that ChatGPT output depends on the prompts given to it, operates in a 'grey area' in terms of accuracy and reliability, and 'cannot substitute either the human intelligence or the humane element' in decision-making

Insights

AI use in Indian courts, including the government-funded e-Courts project and translation and transcription services like Nyaay Ai and Adalat AI, is currently focused on improving efficiency and reducing time spent on routine tasks. As at March 2026, India’s judicial system faces a backlog of over 50 million cases, which would take human judges more than 300 years to resolve at the current pace

Information uploaded as at March 2026

AT A GLANCE

AI is increasingly used across the justice system, particularly by law enforcement for data analysis, crime detection, surveillance, and predictive policing. Tools such as facial recognition and AI-enhanced CCTV analysis (e.g., AMPED FIVE) improve investigations and evidence handling. Prosecutors employ AI to support case management, including filing first information reports and analysing witness testimony. Courts use AI under the e-Courts project for automated case handling, legal research, translation (SUVAS), and summarisation and legal research (SUPACE), while platforms like Nyaay AI and Adalat AI enhance efficiency and accessibility. Victims have access to specialised chatbots for case updates, summarisation, and translation. Training programmes for judges and police are expanding AI literacy to strengthen justice delivery.

India does not have dedicated AI legislation. Local policies and guidelines for lawyers and judges are emerging, and general rules of criminal procedure also apply. There are plans to amend amend existing laws, rely on sectoral regulations and adopt voluntary frameworks, as opposed to enacting new pieces of legislation specifically for governing AI in India.

USE

Law enforcement

Operational support

Uttar Pradesh Police have also launched an AI-powered legal assistance platform called ‘NyaySanhita.in’, which allows users, especially women, to describe incidents in Hindi or English and receive guidance on applicable legal provisions, explanations, and links to the new criminal codes to make justice more accessible. The tool is intended to help women understand their rights, report crimes more effectively, and reduce errors in legal application, while also assisting police and lawyers in building stronger cases.

In April 2026, the Central Bureau of investigation (CBI) has created an AI-powered chatbot named ‘Abhay’ to assist in verifying digital arrest scams. It allows users to verify whether any online notices they receive are authentically issued by the CB

Data review and analysis

In a press release dated 25 February 2025, the Indian Ministry of Law and Justice confirmed that AI is being integrated into policing and law enforcement to enhance crime detection, surveillance, and criminal investigations. Some examples of such applications are surveillance and investigation by way of facial recognition systems with databases and AI-powered forensic analysis for examination of evidence. Delhi Police have been using AI-based image enhancement and facial recognition tools to aid investigations. For example, the AMPED FIVE system is used to clarify low-quality CCTV footage. These enhanced images are then matched against various government and private databases, including driving licences, voter rolls, and vehicle records.

According to the Ministry of Law and Justice, AI is also being used for improving crime tracking and intelligence systems such as Crime and Criminal Tracking Network Systems (CCTNS) and integration with India's e-Prisons database (for digital management and tracking of information about inmates and prisons) and e-Forensics database (for storing, managing and sharing forensic evidence data and laboratory reports electronically).

In Delhi, AI is being deployed to modernise the capital’s security architecture, as part of the Safe City Project. An AI-powered surveillance network is integrated through a new Command, Control, Communication and Computer Centre. The network is being deployed across 10,000 new CCTV cameras and 15,000 existing cameras, enabling real-time video analytics and monitoring to detect and respond to threats more rapidly and effectively.

The Delhi Traffic Police are also planning to implement an AI-powered Integrated Traffic Management System, that uses AI to detect and manage traffic violations and improve overall traffic flow across the city. The system will deploy smart adaptive traffic signals and automatic number plate recognition cameras which will monitor roads in real time, identifying vehicles breaking rules and automatically generating traffic files without human intervention.

Predictive analytics

The Ministry of Law and Justice has confirmed that AI is used in India for predictive policing by analysing crime patterns, high-risk areas, and criminal behaviour, enabling law enforcement to take proactive measures.

Prosecutors

Case management

In its press release dated 25 February 2025, the Ministry of Law and Justice also confirmed that AI is being used for assistance in filing first information reports, and for tools to examine and enhance witness testimony and courtroom evidence.

Courts

As at March 2026, AI use in Indian courts is focused on improving efficiency and reducing time spent on routine tasks. In India’s judiciary, there are over 50 million cases backlogged (including but not limited to criminal matters), which would take human judges more than 300 years to resolve at the current pace.

Case management

India’s e-Courts initiative, launched in 2004 and spearheaded by the Supreme Court of India, is a government-funded nationwide project incorporating AI for automated case management, legal research, and document digitisation, aiming to reduce case backlogs and improve judicial efficiency (including in criminal proceedings). The project is being implemented under the joint partnership of the Department of Justice, the Ministry of Law and Justice, the Government of India and the Supreme Court of India’s e-Committee in a centralised manner through the respective High Courts.

Phase I	Phase I of the project was approved in 2010 and enabled computerisation of 14,249 district and subordinate courts by 2015
Phase II	Phase II was commissioned in 2015 and involved enhancing judicial services for litigants and lawyers by providing them with technology-enabled infrastructure. It involves, inter alia, improved ICT infrastructure and video conferencing
Phase III	Phase III was approved in 2023, with the main objective of creating a unified technology platform for the judiciary. AI technologies and blockchain are being incorporated across High Courts in India up to 2027. A formal data-governance framework is required for all AI technologies being used in e-Courts software applications. As at March 2026, these projects are operational but are yet to realise full implementation

Potential AI-driven enhancements involved in Phase III include:

Intelligent scheduling and case timeline prediction: AI-powered tools are deployed to optimise court scheduling, reduce delays, and ensure timely hearings. Predictive analytics can assess case timelines, offering valuable insights into probable outcomes based on historical data.
Automated filing and case management: AI-driven document automation and OCR technologies are enhancing the accuracy and speed of filing legal documents. These systems minimise manual data entry, improving efficiency and reducing administrative burdens on court staff.
Natural language processing for legal research and translation: natural language processing tools revolutionise legal research by providing judges and lawyers quick access to relevant precedents and judgments. Additionally, AI-based translation systems facilitate multilingual access to legal resources.
Enhanced case information system and chatbots: AI-powered chatbots are deployed to assist litigants by providing real-time case updates and procedural guidance. It reduces the dependency on court officials for basic queries and improves use engagement with the judicial system.
Data security and privacy measures: acknowledging the imperative of data protection, a sub-committee composed of High Court judges and technical specialists has been constituted to guarantee secure connectivity and authentication protocols. This committee evaluates and enhances the e-Courts project’s digital infrastructure to protect citizen privacy.

The Supreme Court of India has also piloted two AI initiatives:

SUVAS

The Supreme Court Vidhik Anuvaad Software ('SUVAS') is an AI tool for translating documents and orders from English into regional languages, such as Hindi, Punjabi and Gujarati. Launched in November 2019, it uses machine assisted translation, trained by AI.

'SUVAS' has been used to translate criminal judgments and, overall, has already translated over 31,184 Supreme Court judgments into 16 regional languages, including Hindi, Tamil, Marathi, Bengali, and Kannada. Translated rulings carry a disclaimer that relieves the Supreme Court of all mistakes in it.

While developing the 'SUVAS' software, the Supreme Court faced a problem in the lack of unified vocabulary for legal jargons in regional languages which led to inaccuracies in the translation process. In response, the Bar Council of India has been working on developing a legal vocabulary that will be used across all courts in India for translation in several regional languages.

SUPACE

The Supreme Court Portal for Assistance in Court Efficiency ('SUPACE') is an AI-driven tool assisting judges by summarising case files. The system processes facts and makes them available to judges looking for an input for a decision. 'SUPACE' operates in four parts:

File preview: converting case files (typically available as PDFs) into text, with a search tool to browse through all files.
Chatbot: Text and voice-enabled chatbot helps to give a quick overview of the case. In a matter of minutes, the bot suggests further questions to be asked for better understanding, and the entire question summary can be printed by the user.
Logic gate: a fact extraction system for the chatbot is divided into four parts: Synopsis, FAQs, Evidence, and Case Law. These give information about the case such as overview, chronology, and judgment.
Notebook: an integrated word processor which truly makes the tool an end-to-end system. A summary of the case can be prepared by simply collating all information auto-extracted from the database using AI. In addition, voice dictation can be used to prepare notes on the drafting tool.

Moreover, the Supreme Court is testing (with IIT Madras) prototypes of AI tools for resolving filing defects, data and metadata extraction, and such tools will be integrated with the electronic filing module and the case management software for smart scheduling and backlog reduction.

‘Nyaay AI’ is an AI platform used by the Indian Supreme Court and 16 of India’s 25 High Courts and co-founded by PanScience Innovations. The platform is a commercial tool and was co-created with the judiciary to ensure alignment with real judicial needs.

Our suite of products spans defect detection, e-filing automation, case clustering, metadata extraction, bench allocation, headnote generation, live transcription, multilingual translation, and judgment analysis—all designed to make courts faster, more efficient, and more accessible.

Anshul Pandey, founder and CEO of PanScience Innovations, August 2025

‘Adalat AI’ is a legal-tech nonprofit, which provides AI-based real-time transcription in Indian courts, converting court proceedings (witness statements, cross-examinations, judgments, orders) into text in real time, assisting or augmenting the work of human stenographers. It also offers translation support, enabling the translation of statements and judgments into regional languages to ensure participants understand. As at March 2026, Adalat AI has been integrated in over 4,000 courts across nine Indian states, with the aim to expand to 7,500 active courtrooms by 2027. Some courts have reported that Adalat AI has reduced case timelines by 30-50% by alleviating pressure on overburdened courts.

A crucial aspect of the criminal justice system is ensuring that case papers are easily accessible in multiple languages. The accused may not know English, or even Hindi, but could speak one of India’s many other languages or dialects. For justice to be meaningful, documents must be available in a language the individual understands. This requires a strong, well-designed system. Unfortunately, the current framework is outdated — it hasn’t been updated for seven or eight years — while technology has advanced rapidly. This gap calls for a serious re-examination, especially since, based on personal experience, the existing system still falls short.

Justice Gautam Patel, High Court of Bombay, September 2025

Legal research, analysis and drafting support

The SUPACE system piloted in the Supreme Court (discussed above) is an AI portal designed to make research easier for judges by highlighting relevant legal precedents to the judge user. The former Chief Justice of India has confirmed that the tool will not spill over into judicial decision-making.

India’s National Informatics Centre has also developed a retrieval-augmented-generation-based judicial search assistant for Supreme Court justices (a RAG-based model combines information retrieval with language generation to produce more accurate answers). It will serve as a specialised search service to help judges quickly find specific information. The tool will first be available to all of India’s Supreme Court judges with plans to scale the judicial assistant to the entire system.

Judges in India have voluntarily disclosed usage of AI tools. There are examples of judges using ChatGPT, clarifying that this was not done to guide judicial opinion or delegate decision-making. These are discussed below (see ‘Cases’).

Decision-making support

India’s government has announced plans to infuse AI in the judiciary to expedite verdicts in ‘petty crime’ and traffic cases, introducing ‘robo judges’, whereby human judges will use AI to process case information, background details, and past orders to help judges deliver verdicts faster.

Defence

In a survey conducted by Manupatra Academy between 9 May 2015 and 19 May 2025, it was revealed that more than half of the legal professionals surveyed (including those working on criminal cases) currently utilise AI tools in operations, while 73.7% have interacted with generative AI applications for their work. However, more than half of the surveyed users noted barriers to widespread adoption of AI in legal enterprises, particularly data privacy and the security of sensitive client information. Users also expressed reservations about the quality and reliability of AI-generated content.

Administrative support

‘Lucio’ is an AI-powered legal assistant that helps law firms, in-house teams, and researchers extract, analyse and manage legal documents efficiently. Lucio has been integrated into the Bar Council of Delhi’s e-research library, launched in July 2025.

Victims

Under Indian criminal procedure law, victims of criminal offences have standing in criminal proceedings (Jagjeet Singh v. Ashish Mishra, Monu Criminal Appeal 632 of 2022 (18 April 2022)).

Case management

In March 2026, Adalat AI launched a WhatsApp-based national case information helpline in partnership with Anthropic. The service enables users to obtain real-time case status and history, receive AI-generated summaries of public court documents, access translation and interactive querying functions in Indian languages, and sign up for WhatsApp alerts regarding case updates. Reportings on the Kerala rollout that Adalat AI is being used or mandated in trial courts for recording witness statements and evidence, and that it is being piloted in matters including atrocity trials and sexual violence cases.

TRAINING

AI training is being dispensed to members of the Indian judiciary. In the May 2025 survey by the Manupatra Academy (see above under 'Defence'), 88.6% of the legal professionals surveyed believed that AI training is crucial.

In December 2024, the National Judicial Academy in Bhopal hosted a workshop for judges of various High Courts across India on information and communication technologies, including dedicated sessions on the use of AI in the courtroom.

Moreover, from April 2025 judicial officers presiding over trial courts began receiving training on the use of AI tools to aid and expedite judicial decision-making, particularly in ‘routine cases’. While no official reports are available, the press reporting on these training efforts clarifies that the training pertains to the use of AI tools to process case information, background details, and past orders towards greater efficiency, rather than in an effort to augment or replace judicial decision-making. As at March 2026, there is no publicly available material to suggest dedicated judicial training on deepfake detection exists in India.

Police training institutions have also begun integrating AI into their professional development programmes. At the Police Training Centre in Medchal, Hyderabad, officers attended a refresher course in August 2025 led by an AI expert. The session ‘offered invaluable hands-on and insightful training on how AI can be leveraged in modern policing’. Training focused on phishing and scam detection, deepfake identification, social media analysis, mobile tracking, digital evidence collection, and voice recognition.

For lawyers, AI training programmes are being offered by private educational or training organisations in partnership with or certified by governmental authorities. For example:

The 'Artificial Intelligence, Law and Justice’ course was offered by IIT Madras in coordination with the National Programme on Technology Enhanced Learning under the aegis of the Ministry of Human Resource Development and the Government of India's ‘Swayam’ initiative.
'LawSikho' is a certificate programme jointly offered by LawSikho, a private training company, NALSAR University, and the National Skill Development Corporation, a company set up as a public-private partnership under the aegis of the Ministry of Skill Development and Entrepreneurship, Government of India.

Moreover, in May 2024, the Bar Council of India directed that all centres of legal education regulated by the Council incorporate “modern technologies”; including AI into their curriculum.

Today’s police force must be as smart as the criminals they face. Our constables and officers must be equipped not only with weapons and discipline but also with awareness and digital intelligence. Training like this prepares them for real-world challenges

Nikeelu Gunda, AI expert delivering police trainings in India, August 2025

REGULATION

The only AI specific regulation in India is the 2026 amendment to the Information Technology Act, 2000 (see ‘Data Protection Legislation’ below). In addition, some sectoral regulations and guidance have been issued. For example, the Union Ministry of Electronics and Information Technology has issued guidance to platforms and intermediaries when deploying or developing AI in a 2024 advisory. There is no general statute on the use of AI tools in court, but local guidelines for lawyers are beginning to emerge. India has not formally adopted or endorsed the UNESCO Guidelines for the use of AI systems in courts and tribunals (2025) although judicial guidance, including the Supreme Court’s 2025 White Paper , is broadly consistent with the Guidelines on themes such as human oversight, confidentiality, and AI being an assistive rather than a substitutive tool.

Guidelines for practitioners

Judges and lawyers are bound by their general ethical and professional obligations, which may apply to the use of AI tools even if AI is not mentioned. For example, the Bar Council of India has issued Rules of Professional Standards which include the duties advocates owe to the court and to the client. In addition, local guidelines on the use of AI in court are beginning to emerge.

Bombay Bar Association: Guidelines on the use of AI by the judiciary

In early July 2025, the Bombay Bar Association issued guidelines on the use of AI to its members, described as ‘the first of their kind to govern the ethical and practical use of artificial intelligence in the judicial process’. The Association’s ‘practical guidelines’ require its members independently to verify every ‘output generated by an AI tool’ before relying on it or sharing it ‘with any Court or client’. Members must refrain from ‘providing client-related data into AI tools’ and assess the ‘risk of exposure’ of ‘sensitive or privileged information’. The Association recommends that they seek a client’s written consent before using AI tools to process the client’s information and maintain records of any ‘interactions with AI tools’.

State of Kerala: Policy Regarding Use of Artificial Intelligence Tools in District Judiciary

In late July 2025, the Kerala High Court became the first state in India to introduce a Policy Regarding Use of Artificial Intelligence Tools in District Judiciary. The policy applies to ‘the District Judiciary in Kerala’ and the ‘employees assisting them in their diverse judicial work’. Violations may result in disciplinary action. The policy warns against the ‘indiscriminate use’ of AI which may result in privacy violations, data security breaches, and the ‘erosion of trust in judicial decision-making’. The policy is aimed at ensuring that ‘AI tools are used only in a responsible manner’ and in compliance with ‘ethical and legal obligations’ to ensure ‘human supervision, transparency, fairness, confidentiality, and accountability’. The policy prohibits the use of AI tools ‘as a substitute for decision-making or legal reasoning’. Only AI tools approved by the High Court of Kerala or the Supreme Court of India may be used. AI may be used ‘solely as an assistive tool’ under strict human supervision. Any results generated by AI tools must be verified by judicial staff, qualified translators or the judges themselves. Courts must keep a ‘detailed audit’ of all instances of the use of AI tools. Members of the judiciary and employees assisting them must participate in training programmes on the ethical, legal, technical and practical aspects of AI.

Criminal procedure rules

Certain aspects of Indian evidentiary and criminal procedure will likely regulate the use of AI in criminal proceedings even though AI is not specifically targeted or mentioned. Some key provisions are set out below.

First, section 63 of the Bharatiya Sakshya Adhiniyam, 2023, an Act ‘to consolidate and to provide for general rules and principles of evidence for fair trial’, directs how ‘electronic records’ are to be proved in evidence. This will likely cover how AI generated materials, server logs, and related matters are required to be proved in both civil and criminal proceedings.

Second, section 72 of the Bharatiya Nyaya Sanhita, 2023, an Act to consolidate and amend the law on criminal offences, prohibits disclosure of the identity of victims of sensitive offences, particularly sexual assault and other serious crimes, to prevent stigma, trauma, or threats to victim safety. Similarly, section 23 of the Protection of Child Sexual Offences Act, 2012 prohibits disclosure of identity (including name, address, school, family details, or photographs) of a victim. These provisions may affect the data sets that can be collated for use with AI tools.

Finally, section 94 of the Bharatiya Nagarik Suraksha Sanhita, 2023, an Act ‘to consolidate and amend the law relating to Criminal Procedure’, authorizes courts or investigating officers to call for the production of a document or other thing in the course of an investigation or criminal proceeding. A document is defined under the Bharitiya Sakhsya Adhiniyam (see above) to include an electronic record. This will therefore likely cover the power of Indian law enforcement agencies and criminal courts to call for the production of records relating to AI platforms.

Data protection legislation

There are various laws that do not directly seek to regulate AI, but may affect the development or use of AI in India. These laws are designed to be technology-agnostic (i.e., the principles in these laws are intended to apply regardless of which technologies are in use). Key examples are set out below.

Digital Personal Data Protection Act, 2023, and Digital Data Protection Rules, 2025

The Digital Personal Data Protection Act, 2023 (‘DPDP Act’) is India’s first comprehensive data protection law. The core operational provisions of the DPDP Act, relating to inter alia:

consent and corresponding aspects;
obligations applicable to data fiduciaries; and
obligations applicable to Significant Data Fiduciaries (entities classified according to factors such as nature and volume of personal data processed) (SDFs),

will be implemented on 13 May 2027.

On 13 November 2025, the Ministry of Electronics and Information Technology formally brought into force the Digital Personal Data Protection Rules, 2025 (‘DPDP Rules’). The DPDP Rules clarify key implementational aspects of the DPDPA, marking a significant milestone in the rollout of Act.

The DPDP Rules may apply to the use of AI tools in criminal proceedings when AI tools are being used to process digital personal data contained in pleadings, chargesheets, medical records, evidence, etc. Whether the statutory obligations accruing from the DPDP Rules will apply hinges on the identity and function of the parties and whether they are acting as a data principal, data fiduciary, or data processor under the DPDP Rules.

There are some exceptions to the application of the Act. Section 17 of the DPDP Act provides exemptions from certain important provisions such as: section 16, concerning the processing of personal data outside India; specified obligations imposed on data fiduciaries under Chapter II (except sections 8(1) and 8(5)); and the rights and duties of data principals under Chapter III, in specified circumstances. These circumstances include:

processing that is necessary for the enforcement of any legal right or claim;
processing by a court, tribunal or other body in India entrusted by law with judicial, quasi-judicial, regulatory or supervisory functions, where necessary for the discharge of such functions; and
processing undertaken in the interest of prevention, detection, investigation or prosecution of any offence or contravention of any law.

Accordingly, to the extent that personal data is processed in connection with the functions contemplated under section 17, the DPDP Act provides a qualified exemption from several key obligations and provisions.

As per the DPDP Rules, SDFs must undertake due diligence to ensure that any technical measures including the algorithmic software used for handling personal data are designed and verified to safeguard against any risks to the rights of data principals.

In the event of a failure by an AI service provider to maintain reasonable security practices in respect of the data collected or processed by it, such service provider may be subject to a fine, which may extend to INR 50 Crore (approximately USD 5.38 million at March 2026 exchange rates).

Notably, the existing data protection framework under section 43A of the Information Technology Act, 2000, read with the Information Technology (Reasonable Security Practices and Procedures and Sensitive Personal Data or Information) Rules, 2011, will remain in force and not be repealed until the end of the 18-month period of implementation, i.e. 13 May 2027. And the enforcement of the DPDPA will not impact the applicability of Section 72A of the IT Act and corresponding penalties for the wrongful disclosure of personal data of an individual, in breach of consent, or a lawful contract with such individual (see further below). Information Technology Act, 2000

In addition, any unauthorized access to user data uploaded on an AI platform (by third parties) and/or illegal dissemination of such data by the AI platform, may have legal consequences under the Information Technology Act, 2000 (’IT Act’). It includes , including the following provisions:

Section 43	Imposes a civil penalty (requiring payment of compensation to a victim) for damage to computer systems, including unauthorised access, data theft, introducing computer viruses, disrupting systems, or denying access to authorised users
Section 43A	Holds bodies corporate liable to pay compensation for failure to protect sensitive personal data, where negligence in implementing security practices causes wrongful loss or gain
Section 66	Criminalises the act of any person fraudulently committing the acts covered under section 43 (above) and prescribes a jail term of up to 3 years and/or a fine
Section 72	Penalises any person who, having access to electronic records or information by virtue of powers under the Act, discloses such information without consent, with imprisonment of up to two years or fine
Section 72A	Provides punishment for disclosure of information in breach of lawful contract. Any person, including an intermediary, who secures access to personal information while providing services under a contract and discloses it without consent, causing or likely to cause wrongful loss or gain, is punishable with a fine

Information Technology (Sensitive Personal Data and Information) Rules, 2011

In respect of data privacy, the Information Technology (Sensitive Personal Data and Information) Rules, 2011 (‘SPDI Rules’) issued under the IT Act govern the collection and disclosure of sensitive personal data and information, defined to include biometric data (including facial patterns, fingerprints, etc.) and financial information. While the SPDI Rules apply to ‘bodies corporate’, who are collecting sensitive personal data and information, and may not apply to the collection of such data by law enforcement agencies, in the event that third party services are used by law enforcement agencies to process or analyse sensitive personal data or information, and particularly where such third parties store the sensitive personal data or information on their own servers, the SPDI Rules may govern the collection and/or processing of such data.

Amendment to the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021

The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 (‘Guidelines’) implement provisions of the IT Act, and regulate the actions of intermediaries (entities receiving, storing or transmitting data) operating in the cyber space. All websites operating in India are required to comply with these Guidelines, and failure to do so can expose these websites to prosecution for any acts committed, even by third parties, using these websites. It is only when intermediaries comply with the Guidelines that they can claim safe harbour from offences committed using their platforms by third parties. This is specifically prescribed under section 79 of the IT Act.

Amongst other things, the Guidelines require intermediaries (websites and other online players) to ensure that harmful information is not uploaded on their platforms, or, if uploaded, is promptly taken down on government/court orders, or even voluntarily, by these intermediaries.

On 20 February 2026, the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, came into effect. The 2026 amendment deals with synthetic and AI-generated content (including creation and circulation of deepfake videos), introducing new compliance duties for platforms and stricter timelines for takedown/ handling of such content by intermediaries. The salient features of the 2026 amendment are:

Amendment to the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021

Rule 2(1)(wa)	This provision defines the term 'Synthetically Generated Information' (SGI) to include any artificially created information which appears to be authentic or true i.e. information created by AI which can be projected as a recording of real events. The proviso to Rule 2(1)(wa) excludes certain usage from the ambit of SGI, such as routine or good faith editing (formatting and technical corrections), routine or good faith creation of content (educational material), and enhancing the quality or accessibility (captions, translations, transcripts).
Rule 2(1)(A)	The definition of the term 'information' used in the Guidelines has been revised to include SGI. This means that obligations placed upon intermediaries in relation to information (including taking down harmful content, taking reasonable efforts to prevent posting of harmful content) now extend to SGI or AI created information.
Rule 3(3)	Every intermediary offering a computer resource that enables, permits or facilitates creation/modification/alteration of SGI must deploy reasonable and appropriate technical measures, including automated tools and mechanisms, to prevent creation and dissemination of SGI that violates any applicable law. Any SGI generated from the intermediary’s platform must be prominently labelled or embedded with a permanent unique metadata or identifier, in a manner that is easily noticeable and adequately perceivable.
Rules 3(1)(d); 3(2)(a)(i); 3(2)(b)	The 2026 amendment shortens several timelines applicable to intermediaries, including for takedown of unlawful information upon receipt of actual knowledge, resolution of complaints by a Grievance Redressal Officer from the date of its receipt, expeditious resolution of complaints pertaining to prohibited content listed under Rule 3(1)(b) and removal of content pursuant to a complaint regarding intimate imagery from the time of its receipt.
Rule 4(1A)	In addition to the obligations under Rule 3(3) (above,) social media websites with more than 5 million users (referred to as ‘significant social media intermediaries’) are also required to obtain declarations from their users uploading any information, on whether such information is SGI. The websites have been directed to deploy reasonable means to verify the veracity of these declarations. If a website knowingly permits, promotes, or fails to act against an SGI in breach of the Guidelines, or does not comply with Rule 4(1A), it will be deemed non‑compliant with the due‑diligence requirements under the Guidelines.

Human rights

Domestic and international human rights instruments may be relevant to the use of AI tools in court. For example, article 14 of the Indian Constitution guarantees equality before the law and article 21 protects a person’s life and liberty. Indian courts have recognised that the right to privacy flows from the protections to life and liberty under article 21 of the Indian Constitution. At the same time, articles 14 and 17 of the International Covenant on Civil and Political Rights and articles 16 and 40 of the Convention on the Rights of the Child, to both of which India has acceded, protect the right to a fair trial and the right to privacy.

Outlook

As regards the approach to AI regulation in general, in November 2025, the Ministry of Electronics and Information Technology issued the India AI Governance Guidelines setting out India’s overarching framework for the development and deployment of trusted AI systems.

The AI Governance Guidelines clarify the approach that the Government is expected to take in relation to AI adoption and risk mitigation in India. The intent is clear that 一 as a short and mid-term strategy一the Government is leaning towards amending existing laws, relying on sectoral regulations and the adoption of voluntary frameworks as opposed to enacting new pieces of legislation for governing AI in India. The Governance Guidelines envisage the adaptation of existing laws to cover AI regulation, recommend voluntary, risk proportionate measures, and emphasise the establishment of accessible grievance redressal mechanisms. They propose the strengthening of institutional capacity, including enabling sectoral regulators to take the lead in their respective domains, as well as the introduction of training programmes and techno-legal measures to support responsible AI.The Governance Guidelines undertake a systematic assessment of legal gaps within the existing framework and propose targeted amendments to statutes, including the IT Act , the Indian Copyright Act, 1957, and the ‘DPDP Act’ (see ‘Data Protection Legislation’ above).

With respect to the legal sector, the Research Unit of the Ministry of Justice concluded in a February 2025 press release on Digital Transformation of Justice: Integrating AI in India's Judiciary and Law Enforcement that ‘responsible AI adoption requires strong data security, legal reforms, and transparency to ensure it supports rather than replaces human judgment in judicial processes’ and the future of AI in the legal sector would be ‘shaped by AI-powered legal research, blockchain-secured case records, judicial transparency through AI analytics, and enhanced cybersecurity in law enforcement’.

The Supreme Court of India, through its Centre for Research and Planning (CRP) released a White Paper on Artificial Intelligence and Judiciary in November 2025. The CRP’s proposals include:

institutional safeguards for ethical use of AI such as establishment of an AI Ethics Committees within courts, with the technical expertise to identify low-risk tasks suitable for AI and create policies governing responsible use, deployment of secure in-house AI tools rather than reliance on external public tools, policy for ethical use of AI, disclosure requirements, and training on ethical use of AI;
demarcating the possible uses of AI in the courts for administrative and assistive tasks rather than adjudication (summarisation, drafting support, transcription, translation, calculations, case management, etc.);
guidelines for the responsible use of AI in judicial institutions
creation of suggestive guidelines for the responsible use of AI for lawyers; and
creation of guidelines for the responsible use of AI for law clerks.

CASES

The use of AI tools has been discussed in a number of cases before the Indian courts.

Judicial use of AI

In some instances, AI tools have been consulted in the decision-making process. For example, in the judgment of Jaswinder Singh v. State of Punjab, CRM-M-22496-2022 (27 March 2023), the Punjab and Haryana High Court asked ChatGPT for feedback to gain a more comprehensive understanding of whether bail should be granted in cases in which the alleged crime involved ‘cruelty’. The judge clarified that this reference was solely intended to provide a broader understanding of bail jurisprudence without dictating the merits of the case. Similarly, in the case of Md Zakir Hussain v. State of Manipur, WP(C) No. 70 of 2023 (23 May 2024) the Manipur High Court used ChatGPT for research - but did not delegate the decision-making - in a service matter when the state government failed to furnish the judge with essential information on the service rules of Village Defence Force personnel.

In Christian Louboutin SAS v. The Shoe Boutique, CS(COMM) 583/2023 (22 August 2023), the plaintiffs relied on a ChatGPT response affirming that they were known for their iconic, red-soled shoe to support an argument that the defendant had breached trademark restrictions. The Delhi High Court found in favour of the plaintiff but rejected the ChatGPT response.

[T]he responses from ChatGPT cannot be the basis of adjudication of legal or factual issues in a court of law. The response of a Large Language Model (LLM) based chatbots such as ChatGPT depends upon a host of factors including the nature and structure of the query put by the user, the training data etc. Further, there are possibilities of incorrect responses, fictional case laws, imaginative data etc. generated by AI chatbots. Accuracy and reliability of AI generated data is still in the grey area. At the present stage of technological development, AI cannot substitute either the human intelligence or the humane element in the adjudicatory process. At best the tool could be utilised for a preliminary understanding or for preliminary research and nothing more.

Prathiba M. Sing, Judge, Delhi High Court, 2023

Misuse of AI in court filings

Indian courts have also been confronted with cases of AI hallucination in legal proceedings. In Gummadi Usha Rani v. Sure Mallikarjuna Rao, 2026 SCC OnLine SC 341 (27 February 2026), arising out of an impugned final judgment passed by the High Court of Andhra Pradesh at Amravati, the Supreme Court took cognizance of an order passed by the Andhra Pradesh trial court which relied on fabricated AI generated citations. The trial court dismissed the objections raised by the defendants by placing reliance on non-existing precedents. The Supreme Court noted that ‘a decision based on such non-existent and fake alleged judgments is not an error in the decision making’ and treated the citation of AI-generated fake case law in court proceedings as an issue of serious institutional concern. The Supreme Court underscored that neither judges nor lawyers can rely on AI-generated citations without independent verification and indicated that such conduct may carry misconduct implications.

In Deepak Rakheja and Anr v. Omkara Assets Reconstruction Private Limited, Civil Appeal No. 12195/2025, arising out of insolvency proceedings initiated by Omkara Assets Reconstruction Private Limited against Gstaad Hotels, the Supreme Court uncovered a filing (rejoinder) containing numerous fabricated and fictitious authorities generated with the assistance of AI tools. The Supreme Court treated the incident seriously because it involved not merely inaccurate citation, but the introduction of non-existent precedents into the judicial record. These observations, reported widely in legal and mainstream media outlets, were made by the Bench during the course of the hearing, but were not recorded in a substantive order, and no formal sanctions or other measures have been ordered pursuant thereto. However, in the (widely reported) courtroom exchange the Bench observed that if the citations were indeed fictitious or AI generated, the Court “would put the Appellant to task”.

In Greenopolis Welfare Association v. Narendra Singh and Ors, CM(M) 1909/2025, CM APPL. 61372/2025 & CM APPL. 61371/2025 (25 September 2025), a similar concern was identified by the Delhi High Court. The petition was found to cite fabricated and non-existing judgments, including fabricated portions and excerpts. The petition was therefore dismissed as withdrawn, at the petitioner's request.

Deepfakes and synthetic media

Indian courts are starting to hear disputes involving deepfakes and AI manipulated evidence. But these remain limited and showcase emerging judicial responses rather than settled authorities of admissibility rulings on deepfake evidence. In Nirmaan Malhotra v. Tushita Kaul, MAT. App.(F.C.) 180/2024 and CM Nos. 32316-18/2024 (28 May 2024), a Division Bench of the Delhi High Court refused to rely on photographs produced by a husband to allege adultery by his wife in maintenance proceedings under the Hindu Marriage Act, 1955. The court observed that ‘we are living in the era of deepfakes’ and held that such visual material cannot be presumed authentic and must be proved by evidence before the Family Court.

In Gautam Gambhir v. Ashok Kumar/ John Does and Ors, CS (COMM) 287/2026, while not pertaining directly to deepfake evidence, Gautam Gambhir, a former international cricketer and the current head coach of the Indian men's national cricket team, started a civil case before the Delhi High Court seeking relief against the systematic misuse of his identity through AI-generated deepfakes and face-swapped videos. The proceedings were brought in the context of identity and personality rights protection, but they demonstrate the increasing likelihood that courts will be called upon to assess the authenticity of AI-generated media.