How Nyay Apex uses AI

Where AI is used — and where it is not

AI (Anthropic's Claude models) is used for research answers, briefs of a record, first drafts, follow-up tasks from a hearing note you ask it to read, and short courtroom answers. Text extraction, chronology, citation detection, cause-list matching, format checks, limitation and the citation graph run on our own server without AI.

What every AI answer carries

• A visible label that it is AI-generated and must be verified.
• The sources it was given — judgment, citation and page — so it can be checked against the original.
• An entry in the chamber's AI register: feature, model, the question, the answer, its sources, and who verified it. The register can be exported, and a disclosure report for a filing can be generated on request.

What we do not do

• No prediction of the outcome of a case and no scoring of a matter's chances.
• No profiling of judges: where the bench is shown, it is only to find judgments they sat in.
• No automatic filing. A paper-book or draft is a draft until an advocate approves it; drafting exports say so on every page until all checks pass.
• Student accounts produce training drafts only.

Your data and the AI provider

• Only the passages needed for a question are sent, not the whole record.
• Phone numbers, email addresses, Aadhaar and PAN numbers are replaced with placeholders before text leaves our server and restored in the answer.
• The AI provider processes requests outside India. If your matter must not leave India, use the tools that run without AI.

When AI gets it wrong

Every AI answer has a "Report" button. Reports go to an incident register with the cause, consequences and the remedy taken, and are reviewed by the platform team.

Questions

If you have a question about how AI is used on your matters, contact us.