Discuss the role of AI-assisted translation (e.g., SUVAS) in democratising judicial outcomes. What safeguards are needed to keep legal translations accurate and authoritative?

Under Article 348(1)(a), proceedings in the Supreme Court and every High Court must be in English [1], a language most litigants cannot read. AI-assisted translation, led by SUVAS (Supreme Court Vidhik Anuvaad Software), narrows this gap without amending the Constitution. It has widened access to court outcomes, but not yet to justice itself.

Role in democratising judicial outcomes

• Scale: SUVAS was built with MeitY support. By August 2024 it had translated 36,271 SC judgments into Hindi and 17,142 into 16 other languages [2]. It now covers 18 Indian languages [3].
• Free access: Translated judgments are hosted on the e-SCR portal, which is free to search, read and download [3].
• Legal literacy: Under Art. 141, Supreme Court law binds every court. Translation lets citizens and lower-court lawyers read that law in their own language, so they depend less on intermediaries.
• Federal balance: Adding many languages instead of swapping English for Hindi respects linguistic diversity.

Limitations

• No legal force: Only the English text counts in law. Even Hindi judgments of High Courts must carry an English translation under Sec. 7, Official Languages Act [1].
• Access only after the case ends: Judgments are translated, but hearings and pleadings stay in English.
• Uneven spread: Hindi translations outnumber those in all other languages combined [2].
• Terminology risk: Machine translation can blur precise terms such as "acquittal" and "discharge".

Safeguards needed

• Human-in-the-loop vetting: Strengthen the SC and HC AI Translation Committees, which oversee quality and constitutional accuracy [3], and staff them with trained legal translators.
• Certified translations: Give vetted versions limited official standing so litigants can rely on them.
• Standard legal glossaries for each Eighth Schedule language.
• Published language-wise targets to correct the Hindi tilt.
• Data-privacy and bias safeguards through the judiciary's technical sub-committees [3].
• Error-reporting channel and a clear disclaimer on e-SCR.

AI translation has opened judgments to ordinary citizens, but that access is still read-only and uneven across languages. SUVAS should be paired with human certification, shared glossaries and balanced language targets, with AI assisting but never replacing judicial judgment [3]. That would turn translated text into justice litigants actually understand, in keeping with Article 39A's promise of equal access to justice.