SLP Drafting Tool

AI Disclosure Practices

← Back to Home

AI Disclosure Practices

The Supreme Court's AI Committee has not yet notified rules governing the use of AI in legal drafting. The practices below are this tool's own voluntary, best-practice self-disclosure — not statutory compliance — and are designed to be updated once official rules exist.

16

Practices In Place

1

Your Responsibility

AI System Disclosure

AI Language Model

Claude (claude-sonnet-4-6) by Anthropic PBC, USA

Embedding Model

all-MiniLM-L6-v2 (sentence-transformers) — local/on-premise

Knowledge Base

SC Rules 2013 + Practice Manual 2017 + filed SLPs (public domain)

API Type

External cloud API — Anthropic's servers

Training on client data

NO — uploaded documents used for in-context inference ONLY

Output type

Advisory draft — requires mandatory human review

Voluntary Disclosure Checklist

Disclosure on all output documentsIN PLACE

Disclosed via the dedicated AI Disclosure Note (Annexure I) and Transparency Report that accompany every package — not stamped onto the filed document itself, so nothing needs manual removal before filing

No fabricated citationsIN PLACE

Anti-hallucination rules enforced in all AI prompts; uncertain fields highlighted in yellow (<<<HIGHLIGHT>>>)

Mandatory review before downloadIN PLACE

10-item HITL checklist mandatory before output package is unlocked for download

AI system fully documentedIN PLACE

AI Transparency Report auto-generated with every job listing model, KB sources, token counts, highlights

Rests with the human advocateIN PLACE

Advocate certifies professional responsibility in HITL review; AI output is advisory only

Output is never presented as finalIN PLACE

Advisory nature disclosed in the AI Disclosure Note and Transparency Report; HITL review enforces verification before download

DPDPA 2023 alignedIN PLACE

Data Processing Notice shown before upload; 30-day auto-deletion of uploaded files

Used only for SLP drafting assistanceIN PLACE

Tool used exclusively for SLP drafting assistance

HITL required for all outputsIN PLACE

HITL workflow mandatory for all SLP outputs; citations must be independently verified

Lawfully obtained public sourcesIN PLACE

Knowledge base built on SC Rules 2013, Practice Manual 2017 (public domain official publications)

Uploads used for inference onlyIN PLACE

Uploaded documents used for in-context inference only; not stored or used for model training

No prediction of judicial outcomesIN PLACE

Tool is purely assistive; no outcome prediction, risk scoring, or adjudicatory function

AI system fully namedIN PLACE

AI system (Claude, Anthropic) fully disclosed; Transparency Report provides full explainability

Voluntary disclosure documentIN PLACE

Auto-generated with every output package; advocate may sign and file it alongside the SLP

Available if ever requestedIN PLACE

Transparency Report provides a complete audit trail; Annexure I provides a Court-facing note

User's own responsibilityYOUR RESPONSIBILITY

Before using AI-assisted output in court proceedings, confirm acceptability with your Bar / the relevant Court, since no Supreme Court AI Committee rules on AI-assisted drafting have been notified yet

Sensitive judicial dataIN PLACE

Uploaded documents auto-deleted after 30 days by default; data minimisation applied

Before Using Output In Court

No Supreme Court AI Committee rules on the use of AI in legal drafting have been notified as of today. Until such rules exist, it is the Advocate-on-Record's responsibility to independently confirm — with their Bar, the relevant Court, or any Appropriate Authority that may apply — whether any disclosure or approval is expected before using an AI-assisted document in a filing.

This tool generates an AI Disclosure Note (Annexure I) with every output package as a voluntary aid — the advocate may choose to sign and file it alongside the SLP as a matter of professional transparency. It is not a document prescribed by any notified regulation.

Questions or Concerns

The AI Transparency Report (included in every output package) documents the AI system used, the knowledge sources retrieved, and every highlighted field, so that the advocate has a full record of the AI's role in preparing the draft.