English-Only Justice

English-Only Justice

Article 348, Linguistic Access to Courts, and the Limits of AI Translation

Author: Monit Gajjar, Oberoi International School, Mumbai

Executive Summary

This paper identifies Article 348 of the Constitution of India as a structural constraint on linguistic access to justice. While Article 348 permits limited regional-language use in High Courts, Section 7 of the Official Languages Act, 1963 requires authorised regional-language judgments to be accompanied by an English translation. The paper distinguishes between access to judicial information and access to legal authority: although programmes such as SUVAS and e-SCR had translated 83,783 Supreme Court judgments into Indian languages by March 2025, translation alone does not establish co-equal legal authority between language versions.

Using constitutional analysis, government translation data, and comparative evidence from South Africa, Kenya and Canada, this paper examines whether Article 348 contributes to linguistic exclusion and whether AI translation can address this restriction. It argues that technological translation can reduce barriers to comprehension but cannot independently resolve questions of legal authority, accuracy or institutional capacity. It therefore proposes regional-language judicial pilots, co-equal authority for certified language versions, human-reviewed AI translation, standardised legal terminology, and systematic measurement of linguistic access to justice.

Introduction

India’s justice system operates within a country characterised by extensive linguistic diversity. The 2011 Census recorded 128.5 million Indians speaking English as a first, second or third language, equivalent to approximately 10.6% of the population (Census of India, 2011). The Constitution acknowledges this diversity through provisions including Articles 29, 30 and 350A, concerning cultural and linguistic identity and mother-tongue education. At the same time, Article 348 establishes English as the language of proceedings in the Supreme Court and High Courts, subject to specified exceptions.  

This creates a distinction between linguistic recognition and linguistic participation. A citizen may possess constitutional protections relating to linguistic identity while encountering a justice system in which the language used to conduct proceedings and express judicial reasoning is one they do not understand. The question is therefore not merely whether a judgment can be translated after it has been handed down, but whether translation provides the same access to law as participation in the language through which that law is produced and legally applied.

New technological measures have made this issue more pressing. The Supreme Court’s SUVAS system and e-SCR portal have substantially increased the availability of judgments in Indian languages. By March 2025, more than 83,000 Supreme Court judgments had been translated into Hindi and other Indian languages.   These developments can reduce barriers to comprehension. Yet if the legal framework does not clearly establish equivalent authority for different language versions, technology may improve access to judicial information without resolving the underlying institutional structure governing legal authority.

This policy paper therefore asks: to what extent does Article 348 restrict linguistic access to justice, and can technological translation address this restriction without altering the legal framework governing the language in which superior-court law is produced and relied upon?

1. The Constitutional Architecture of Linguistic Exclusion

1.1 Article 348 and the Language Mandate

Article 348(1) provides that proceedings in the Supreme Court and every High Court shall be in English unless Parliament provides otherwise. Article 348(2) permits the Governor of a State, with the previous consent of the President, to authorise Hindi or another official state language for proceedings in the relevant High Court, although the constitutional provision itself excludes judgments, decrees and orders from that authorisation.  

Section 7 of the Official Languages Act, 1963 provides an additional mechanism. With the required authorisation, Hindi or another official state language may be used for High Court judgments, decrees and orders, but any such judgment, decree or order must be accompanied by an English translation issued under the authority of the High Court.  

This distinction is at the heart of the access problem discussed in this paper. The legal framework separates the language that may be used in proceedings from the linguistic requirements governing judicial outputs. A party may therefore participate in proceedings in an authorised regional language while the resulting judgment must also be accompanied by an English translation. The framework provides for multilingual judicial practice, but does not establish a general rule that both language versions possess co-equal legal authority.

Article 348(1) also contains the words “until Parliament by law otherwise provides”, leaving Parliament a constitutional route through which the existing framework could be altered.   The persistence of English therefore presents a contemporary issue: whether the institutional reasons for its privileged position continue to justify the existing framework when regional-language legal terminology, translation infrastructure and technological capacity have developed substantially since the Constitution was adopted.

1.2 A Colonial Inheritance, Not a Neutral Choice

Article 348 did not emerge in a historical vacuum. It developed within a colonial system of judicial administration in which English had become entrenched in the superior courts. As Baxi argues in The Crisis of the Indian Legal System (1982), the institutional development of Indian law cannot be separated entirely from its colonial inheritance.

This inheritance should not be treated as proof that the present framework is inherently illegitimate. Rather, it demonstrates that the privileged position of English was shaped partly by institutional arrangements that predated independent India’s determination of its own linguistic priorities.

This tension was also visible during constitution-making. Members of the Constituent Assembly debated the continued use of English and the practical difficulties associated with replacing it in the legal system. The resulting constitutional settlement attempted to reconcile linguistic change with institutional continuity.

However, historical justification alone cannot determine contemporary policy. The relevant question is whether the practical problems that influenced constitution-making continue to warrant the same institutional arrangement decades later. India’s legal institutions have since had substantial time to develop regional-language terminology, translation systems and technological infrastructure. The basis for retaining English’s privileged position in superior-court law should therefore be evaluated against contemporary institutional capacity rather than treated as inevitable because of its historical origins.

1.3 Tension With Equality, Expression, and Due Process

Article 348 cannot be read in isolation from Part III of the Constitution. Gautam Bhatia describes the Constitution as a transformative document in which equality under Article 14 exerts a “normative influence” across the constitutional architecture (2018). Applied to language and access to justice, this raises a question of substantive rather than merely formal equality: citizens may possess formally equal access to courts while possessing substantially different capacities to understand the language through which legal authority is expressed.

Linguistic access can also be examined through Article 19(1)(a), which protects freedom of speech and expression. This provision does not create an entitlement to obtain judicial decisions in a preferred language. However, it raises a broader question about meaningful participation in public institutions where the language used by those institutions creates significant barriers to understanding.

Article 21 raises a related concern. Its interpretation has placed importance on fair procedure, making meaningful access to legal processes more significant than mere physical or procedural presence. N.R. Madhava Menon’s work similarly emphasises meaningful interaction with legal institutions (Menon, 2002). Linguistic understanding may therefore be regarded as an important component of meaningful access to justice.

These provisions, however, do not establish that Article 348 is unconstitutional or create an express constitutional right for litigants to conduct superior-court proceedings in a language of their choice. Instead, they identify a potential tension between the Constitution’s broader commitments to equality, participation and fair access to legal institutions and the English-language framework established by Article 348.

1.4 Who Pays the Price

The consequences of linguistic barriers are unlikely to be distributed equally. Pierre Bourdieu’s concept of linguistic capital describes language proficiency as a form of social resource: fluency in a language valued by powerful institutions can make those institutions easier to navigate (Bourdieu, 1986).

In India’s legal system, English proficiency can provide advantages in communicating with legal professionals, reading judgments and navigating legal procedures, while people with lower English proficiency may become more dependent on intermediaries.

Language barriers can also compound existing social and economic disadvantages. For communities that already face difficulties accessing legal representation or public institutions, an additional linguistic barrier may further restrict meaningful participation. The policy issue is therefore not simply whether citizens can eventually obtain a translation, but whether the judicial system is sufficiently equipped to prevent language from becoming an additional barrier to engaging with legal institutions.

Contemporary judicial commentary reinforces the relevance of this concern. In his inaugural address at the first N.R. Madhava Menon Memorial Lecture in New Delhi on 17 September 2025, Chief Justice B.R. Gavai stated that “geographical, economic, and linguistic barriers have acted as formidable obstacles, keeping marginalised and vulnerable citizens distant from our courts and legal institutions” (Gavai, 2025). The statement does not establish the size of linguistic exclusion within individual communities, but it demonstrates that language remains recognised as a contemporary access-to-justice issue.

2. Why AI Translation Hasn’t Closed the Gaps

2.1 SUVAS and e-SCR

SUVAS (Vidhik Anuvaad Software) was introduced by the Supreme Court to facilitate translation of judicial documents between English and Indian languages. The expansion of e-SCR has subsequently made translated Supreme Court judgments available through a centralised platform.

The scale of this programme is substantial. As of 28 March 2025, 36,344 Supreme Court judgments had been translated into Hindi and 47,439 into other Indian languages, producing a total of 83,783 translated judgments uploaded to e-SCR with the assistance of High Courts.  

These figures demonstrate significant institutional investment in linguistic accessibility. However, translation volume measures institutional output, not whether litigants actually use, understand or can legally rely upon the translated texts. The number of translations therefore demonstrates an expansion of access to information without, by itself, establishing an equivalent expansion of access to legal authority.

2.2 Accuracy Problems

The limits of translation are also demonstrated by judicial experience.

In Zoharbee & Anr. v. Imam Khan (D) through LRs & Ors., the Supreme Court recorded its dissatisfaction with the manner in which a civil-court judgment had been translated into English. The Court emphasised that precision in legal language matters because individual words and punctuation can affect the understanding of a legal issue.  

This provides a concrete illustration of why translation accuracy is not merely cosmetic. If the meaning of a legal finding changes through translation, the translation can affect how subsequent courts, lawyers or litigants understand the original decision.

The implication for AI translation is not that automated systems are inherently unsuitable. Rather, translation used within a legal system requires verification mechanisms proportionate to the consequences of error. AI can accelerate translation, but accuracy and legal reliability require institutional oversight.

2.3 The Structural Ceiling

The problem therefore extends beyond translation accuracy to the question of legal authority. Article 141 provides that the law declared by the Supreme Court is binding on all courts in India. A translated judgment can improve comprehension without necessarily altering the legal framework governing how that judgment is authoritatively expressed and relied upon.

This creates a two-stage access problem. Can a litigant understand the court’s reasoning?  Can the litigant rely upon the language they understand as an equally authoritative expression of the court’s legal reasoning?

The existing framework does not establish a general rule of co-equal legal authority between English and regional-language versions. Section 7 instead requires an English translation to accompany an authorised regional-language High Court judgment.  

The policy implication is therefore significant. Increasing the number of translated judgments can reduce an information barrier without changing the institutional framework through which judicial authority is expressed. AI can enhance linguistic accessibility, but it cannot by itself determine the legal status of the translation. Any policy seeking to address linguistic exclusion must therefore consider not only how judgments are translated, but also what legal status each language version possesses.

3. Case Studies: Comparative Language-and-Courts Models

The three cases demonstrate different relationships between constitutional language rights and judicial administration.

South Africa demonstrates that constitutional multilingualism can encounter administrative and resource constraints. Kenya demonstrates that constitutional recognition of official languages can coexist with court rules and institutional practices that privilege one language in particular contexts. Canada demonstrates a more developed institutional model in which language rights are supported by judicial capacity, interpretation and rules governing the status and availability of judgments.

3.1 South Africa

Section 6 of South Africa’s 1996 Constitution recognises 12 official languages and requires their status to be treated with parity of esteem. Section 35(3)(k) also provides an accused person with a fair-trial right to a language they understand.

Yet constitutional recognition has not automatically produced multilingual judicial administration. The South African Heads of Court resolution records that, at their March 2017 meeting, the Heads of Court decided that English should be implemented as the language of record in the superior courts in the absence of an executive policy decision.  

South Africa therefore demonstrates that formal constitutional recognition does not automatically translate into multilingual judicial practice. For India, the lesson is that changing the legal status of regional languages would need to be accompanied by administrative capacity, judicial resources and clear rules governing the language of the record.

3.2 Kenya

Article 7(2) of Kenya’s Constitution establishes Kiswahili and English as the official languages of the Republic. Kenya’s Supreme Court Rules, 2020, rule 8, make English the language of the Court, while allowing a party to address the Court in another language provided that seven days’ notice is given to the Registrar (Supreme Court Rules, 2020, r. 8).

This illustrates how constitutional recognition of official languages can coexist with procedural rules that privilege one language within judicial institutions.

Kenya therefore provides a useful case study of the institutional tension between constitutional language recognition and judicial procedure. Its relevance to India lies in the question of whether recognising a language constitutionally is sufficient when the rules governing court proceedings and documents continue to privilege another language.

3.3 Canada

Canada provides a different institutional model. Under the federal Official Languages Act, English and French are official languages of federal courts, and either may be used by a person appearing before a federal court. Federal courts must ensure that judges hearing proceedings can understand the language chosen by the parties without an interpreter.  

The legislation also establishes institutional mechanisms supporting bilingual adjudication. The Government of Canada must consider equal access to justice in both official languages when appointing judges to superior courts, while certain judgments must be made available simultaneously in both languages. Where simultaneous availability is not required, the second language version must be issued at the earliest possible time, and both versions are effective from the time the first version becomes effective.  

Canada’s relevance therefore lies not simply in recognising two official languages, but in institutionalising that recognition. Judicial language capacity, interpretation infrastructure and rules governing judgments demonstrate that multilingual justice requires more than software capable of translating documents. It requires rules governing who may use each language, who must understand it, how judgments are issued and what legal status different language versions possess.

4. Recommended Solutions

The comparative cases reveal a common lesson: legal recognition alone is insufficient.

South Africa demonstrates the implementation difficulties that arise when multilingual constitutional commitments encounter institutional constraints. Kenya demonstrates that constitutional recognition can coexist with procedural rules that restrict practical use. Canada demonstrates the opposite approach: language rights become meaningful when legislation creates corresponding institutional duties concerning judges, proceedings, interpretation and judgments.

For India, this suggests that simply increasing the number of AI-translated judgments would address only the first layer of the problem identified in this paper. The reforms below therefore target four distinct barriers: legal authority, translation accuracy, linguistic consistency and measurable access.

4.1 Activate Article 348's dormant clause. 

Parliament should use the constitutional authority under Article 348 and amend the relevant statutory framework to establish a pilot for regional-language proceedings and judgments with co-equal legal authority in specified High Courts.

The pilot could build upon existing authorisation for regional-language judicial work in selected High Courts. It should establish:

• when regional-language proceedings may be conducted;

• when judgments may be issued in that language;

• how each language version is authenticated;

• which institution certifies translations; and

• what procedure applies when language versions diverge.

The pilot should then be evaluated through measurable indicators including litigant uptake, translation turnaround times, frequency of language discrepancies and the number of disputes concerning differences between language versions.

4.2 Give certified language versions co-equal legal authority

Parliament should amend the relevant statutory framework so that officially certified regional-language judgments within the pilot jurisdictions possess co-equal legal authority with their English counterparts.

The Canadian model demonstrates the institutional value of establishing equivalent authority rather than treating translation as merely an informational service. Canada’s Official Languages Act provides that specified materials issued in both official languages are equally authoritative and establishes detailed rules governing bilingual federal-court judgments.  

For India, the amendment should specify:

1. which institution authenticates each language version;

2. when both versions become legally effective;

3. how discrepancies are identified; and

4. which procedure resolves conflicts between language versions.

4.3 Professionalise legal translation.  

AI translation should be integrated into a certified human-review system rather than treated as a complete substitute for legal translation.

AI could produce the initial translation, after which trained legal-language translators and reviewers would validate the output before certification. The system should also be subject to periodic accuracy audits, with recurring terminology problems and error rates published.

This approach treats AI as an efficiency-enhancing tool while retaining institutional responsibility for the legal reliability of translated judgments.

The need for such oversight is illustrated by Zoharbee v. Imam Khan, where the Supreme Court itself identified problems with the translation of a civil-court judgment.

4.4 Standardise legal terminology across languages.  

The Law Commission of India, in coordination with the judiciary, should lead a multilingual legal-terminology project.

The glossary should:

• establish standard equivalents for recurring legal concepts;

• identify concepts for which no direct linguistic equivalent exists;

• record approved terminology used by courts;

•  be updated as judicial interpretation develops; and

•  be made publicly accessible.

This would reduce the risk that identical legal concepts receive inconsistent translations across judgments and jurisdictions.

4.5 Measure the access gap.  

The Department of Justice should publish disaggregated data comparing litigants’ first language with the language of their proceedings.

The dataset should record:

• whether interpretation was required;

• the language in which proceedings were conducted;

• whether a translated judgment was available;

• the time taken to produce the translation;

• whether the litigant relied upon the translated version; and

• whether disputes concerning translation occurred.

Without such data, policymakers can measure how many judgments are translated but cannot determine whether translation actually improves access to justice.

5. Feasibility and Implementation

The recommendations differ substantially in their legal and administrative demands. They should therefore not be treated as a single reform requiring simultaneous implementation.

The least institutionally disruptive reform is professionalising translation. Existing AI translation infrastructure can be retained while introducing systematic human review, terminology standards and auditing. This could therefore constitute the first stage of implementation.

The second stage should focus on measurement and standardisation. A national dataset and multilingual legal glossary would provide the infrastructure necessary to evaluate whether translation is improving access and to reduce inconsistency across jurisdictions.

The most legally consequential reforms are the proposed regional-language pilot and co-equal legal authority. These require legislative action and coordination between Parliament, the judiciary and participating High Courts. They should therefore begin as limited pilots rather than being introduced nationwide immediately.

The comparative evidence suggests why this phased approach is preferable. South Africa demonstrates that multilingual legal commitments can encounter significant administrative constraints, while Canada’s experience shows that bilingual judicial rights require sustained institutional capacity rather than language recognition alone.

The following framework therefore assesses each recommendation according to the degree of legislative intervention and institutional capacity it would require.

Taken together, the assessment suggests that translation reform, terminology standardisation and access measurement can begin within existing institutional structures, while changes to the legal status of language versions require legislative reform and should therefore be tested through limited pilots first.

6. Conclusion

Article 348 is not simply a neutral administrative arrangement. It establishes English as the default language of proceedings in India’s superior courts while permitting only defined mechanisms for regional-language use. Section 7 of the Official Languages Act subsequently permits authorised regional-language High Court judgments but requires an accompanying English translation.  

The Supreme Court’s translation initiatives represent a substantial institutional response to the resulting information barrier. By March 2025, 83,783 Supreme Court judgments had been translated into Hindi and other Indian languages and uploaded to e-SCR.   Yet the scale of translation does not itself resolve the distinction between understanding judicial reasoning and accessing it through an equally authoritative legal language.

The comparative evidence reinforces this distinction. South Africa demonstrates that constitutional multilingualism can encounter administrative limitations; Kenya illustrates the tension that can emerge between constitutional language recognition and judicial procedure; and Canada demonstrates that bilingual legal authority requires institutional mechanisms governing proceedings, judges, interpretation and judgments.

The proposed reforms therefore treat AI translation as one component of a broader access-to-justice strategy rather than a complete solution. Professional human review would address accuracy; terminology standards would address consistency; systematic data collection would measure actual access; and regional-language pilots would test whether multilingual judicial practice is institutionally viable. Where those pilots demonstrate sufficient institutional capacity, extending co-equal legal authority to certified regional-language judgments would address the remaining distinction between translated information and authoritative law.

The objective is not simply to replace English. It is to ensure that linguistic background does not determine how easily an individual can understand, participate in, and rely upon India’s legal system.

7. Bibliography

Primary legal and institutional sources:

Constitution of India. (1950). Articles 14, 19(1)(a), 21, 141, 343–351.

Official Languages Act, 1963, No. 19 of 1963, s. 7.

Ministry of Law and Justice, Government of India. (2025, April 3). Written reply concerning translation of Supreme Court judgments into Indian languages. Press Information Bureau.  

Supreme Court of India. Zoharbee & Anr. v. Imam Khan (D) through LRs & Ors., 2025 INSC 1245. 

Supreme Court of Kenya. (2020). Supreme Court Rules, 2020, Legal Notice No. 101 of 2020, r. 8. 

Constitution of Kenya. (2010), art. 7.  

Constitution of the Republic of South Africa. (1996), ss. 6, 35(3)(k).

Heads of Court. (2017). Heads of Court Resolution on the Language of Record in South African Courts. Judiciary of South Africa.  

Constitution Act, 1867 (Canada), s. 133.

Canadian Charter of Rights and Freedoms, s. 19.

Official Languages Act, R.S.C. 1985, c. 31 (4th Supp.), ss. 13–20.  

Secondary scholarship:

Austin, G. (1999). Working a Democratic Constitution: The Indian Experience. Oxford University Press.

Baxi, U. (1982). The Crisis of the Indian Legal System. Vikas Publishing.

Bhatia, G. (2018). The Transformative Constitution: A Radical Biography in Nine Acts. HarperCollins India.

Bourdieu, P. (1986). “The Forms of Capital.” In J. Richardson (Ed.), Handbook of Theory and Research for the Sociology of Education, 241–258. Greenwood.

Census of India. (2011). C-17: Population by Bilingualism and Trilingualism, India. Office of the Registrar General & Census Commissioner.  

Constituent Assembly of India. (1949). Constituent Assembly Debates, Vol. IX.

Gavai, B. R. (2025, September 17). Inaugural address at the First Professor N. R. Madhava Menon Memorial Lecture, New Delhi. Reported by LiveLaw.

Mehrotra, R. (2024). Language as a Barrier to Access to Justice: An Analysis in Terms of a Multilingual State. SSRN.

Menon, N. R. M. (2002). Access to Justice and Legal Aid. Oxford University Press.

Seervai, H. M. (2013). Constitutional Law of India (4th ed., Vol. 3). Universal Law Publishing.