In-Draft Compliance Nudge Agent for Sydney Litigation Lawyers: Catch Disclosure, Citation and Privilege Issues Before Lodgement
You’re drafting an expert evidence summary for an Administrative Review Tribunal matter at 6pm. The expert’s report came back with a passage that reads suspiciously polished — and there’s a citation in your own outline you can’t immediately place. The ART’s expert evidence practice direction expects the report and your submission to be accurate and properly attributed. You don’t want a separate verification pass at midnight. You want the editor itself to tell you, while you type, what’s likely to fail review. That’s what the In-Draft Compliance Nudge Agent does.
The problem
Pre-lodgement review in a Sydney litigation practice is typically a check at the end — a partner skim, a paralegal citation pass, sometimes a Bluebook or AGLC4 sweep. By that point the draft is locked in shape and any rework is expensive. Three failure modes have become harder to catch at the end:
- Missing AI disclosure. The ART’s practice directions and the Federal Court’s GPN-AI set expectations around the use of generative AI in proceedings and the practitioner’s responsibility for what they file. If part of a submission, expert statement or witness outline was AI-assisted and disclosure was warranted, that decision needs to be made before the document is finalised — not discovered in cross-examination.
- Dubious citations. Hallucinated authorities (fictitious case names, wrong court, invented paragraph references) are not always obvious on a single read. Catching them at the end of the day, against a deadline, is exactly when they slip through.
- Privilege exposure. Quotes pulled from privileged communications, internal advice, or without-prejudice correspondence sometimes migrate into draft submissions during paragraph reordering. By the time the draft is “finished”, the exposure is buried.
What the In-Draft Compliance Nudge Agent does
The In-Draft Compliance Nudge Agent runs alongside your drafting environment and surfaces three classes of nudge while you write:
- AI disclosure nudges — flags passages whose stylistic and structural markers are consistent with model-generated text where no disclosure has been entered for the document
- Citation nudges — flags every authority reference as you type it, checking the citation against an Australian authority registry (AustLII, Federal Court, ART, NSW Supreme Court) and marking it verified, mismatched or not found
- Privilege nudges — flags passages that resemble content from sources tagged as privileged or without-prejudice in your matter file
Nudges appear inline, are dismissable, and do not block drafting. The point is to move review left — from a midnight verification scramble to a series of small decisions made while the context is still in your head.
How it works
- Tag your matter sources. Upload or link the documents associated with the matter and tag each as privileged, without-prejudice, expert report, or public. The agent uses these tags to drive privilege nudges.
- Draft in your editor. The nudge agent attaches to a local editor session. No draft content leaves your machine.
- Receive inline flags. As you write, citations are checked against the authority registry, passages are scored for AI-likely structure, and quoted strings are matched against your tagged sources for privilege overlap.
- Resolve before lodgement. Each nudge has a recommended action — verify against AustLII, add an AI disclosure entry, re-check the privilege status of a quoted passage. Resolutions are logged.
- Export a readiness report. Before lodgement you export a structured Markdown report of every nudge raised and its resolution, suitable for archiving with the matter file.
Why this matters in Sydney
Sydney litigation teams routinely run matters across the Federal Court (Sydney registry), the ART, the NSW Supreme Court and the Federal Circuit and Family Court. Each forum has its own expectations around expert evidence, AI use and candour to the tribunal. The ART’s Expert Evidence Practice Direction, published with the Tribunal’s other practitioner guidance, sets out how expert evidence is to be prepared and adduced — including the expert’s duty to the Tribunal and the form of expert reports. The Federal Court’s GPN-AI, applicable to matters filed in the Sydney registry, makes clear that practitioners remain responsible for the accuracy of citations and authority references in documents filed, regardless of whether AI was used in their preparation. The Australian Solicitors’ Conduct Rules — Rule 19 (candour to the court) — apply across all of these forums.
The In-Draft Compliance Nudge Agent is the same deterministic verification engine that powers the Citation Verification Agent in RuleCheck by Exegesis (github.com/andrefabre/rulecheck), extended into a live editor integration. Citation checks are deterministic registry queries — not LLM inference — and draft content is not transmitted to external services.
Sources
- Administrative Review Tribunal — Practice Directions and Other Guidance: https://www.art.gov.au/help-and-resources/professionals-and-practitioners/practice-directions-and-other-guidance
- Federal Court of Australia — Use of Generative Artificial Intelligence Practice Note (GPN-AI): https://www.fedcourt.gov.au/law-and-practice/practice-documents/practice-notes/gpn-ai
- Law Council of Australia — Australian Solicitors’ Conduct Rules: https://lawcouncil.au/policy-agenda/regulation-of-the-profession-and-ethics/australian-solicitors-conduct-rules
- AustLII (Australasian Legal Information Institute): https://www.austlii.edu.au/
- RuleCheck by Exegesis (open-source citation verifier): https://github.com/andrefabre/rulecheck
Join the waitlist
The nudge agent is in design partner phase with a small number of Sydney litigation teams. Join the waitlist to be considered for early access and to shape how nudges, disclosure logging and the readiness report work in practice.