Expert Evidence AI Attribution Agent for Melbourne Litigation Lawyers: Prove Which Parts of the Report a Model Touched
Your expert has filed a 40-page report. Counsel asks the question every litigation lawyer in Melbourne now expects in cross: “Dr X, which parts of this report were drafted, summarised or generated with the assistance of an AI tool, and what was the input?” Your expert hesitates. The Federal Court GPN-AI and the Administrative Review Tribunal’s expert evidence practice directions expect a clear answer. The Expert Evidence AI Attribution Agent attaches verifiable attribution metadata to the report before it leaves the expert’s desk, so the answer is already on the page.
Why it matters now
The Federal Court of Australia’s General Practice Note on Artificial Intelligence (GPN-AI) sets expectations that generative AI use in court documents — including expert material — be disclosed and verified. The Administrative Review Tribunal publishes practice directions and guidance for professionals and practitioners which set out how expert evidence should be prepared and presented before the Tribunal. The Australian Solicitors’ Conduct Rules (Rule 19) impose a paramount duty of candour to the court, which extends to the provenance of material an instructing solicitor adduces. Where an expert uses an LLM to draft, summarise, translate or restructure parts of a report and that use is undisclosed — or disclosed without a way to identify which paragraphs were affected — the report becomes vulnerable to a competency or admissibility challenge, regardless of the underlying expertise. The exposure runs both ways: to the expert (whose evidence may be impeached) and to the instructing solicitor (who carries the candour duty).
The 5-minute view
- Expert evidence AI attribution failure is the inability to identify, at paragraph or section level, which parts of an expert report were prepared with AI assistance and which were not
- GPN-AI requires disclosure of generative AI use in material put before the Federal Court
- ART practice directions set out preparation standards for expert evidence before the Administrative Review Tribunal
- ASCR Rule 19 (candour) places the responsibility for the integrity of adduced material on the instructing solicitor
- Verbal disclaimers (“AI was used in places”) are insufficient — they cannot be tested in cross and they cannot survive document handling
- The Expert Evidence AI Attribution Agent attaches structured attribution metadata to each section of a draft expert report
- Metadata records: which sections were AI-assisted, the model/tool category, the human review status, and a hash of the underlying source material
- The agent runs as part of the expert’s pre-finalisation workflow; the metadata travels with the document
What Exegesis is building
The Expert Evidence AI Attribution Agent is a T2 service shape in the Exegesis Legal stack. It attaches verifiable attribution metadata to expert evidence outputs so that an expert report’s AI provenance is recorded at the section level and can be inspected, exported or produced on request. The agent does not generate expert opinion. It does not rewrite the report. It sits alongside the expert’s drafting tool, classifies each section by the nature of any AI assistance disclosed by the expert (drafting, summarising, structuring, translation, none), and emits a structured attribution block — both human-readable (an appendix the expert signs) and machine-readable (embedded metadata) — that an instructing solicitor can include in the brief without re-engineering. The architecture is deliberately narrow: no opinion generation, no transmission of report content to external LLMs, retention only as configured.
The deliverable
- A per-section attribution map of the expert report (paragraph or heading granularity)
- A signed attribution appendix suitable for inclusion in the report itself
- Machine-readable metadata (JSON sidecar) recording assistance category, tool category, and human-review status per section
- A hash record of source materials the expert relied on, so later challenges to provenance can be answered
- An audit log entry suitable for the matter file and for production if attribution is challenged
- Delivered as a pre-finalisation step in the expert’s workflow; integrates with the brief without changing the expert’s drafting tool
Why this matters in Melbourne
Melbourne litigation lawyers run a heavy mix of Federal Court, Supreme Court of Victoria and ART matters — and the expert pool in this city is shared across all three. An expert who is engaged on a Federal Court matter on Monday and an ART matter on Wednesday is operating under overlapping but not identical AI-disclosure expectations. A Melbourne instructing solicitor inheriting an expert’s draft does not have time to interview the expert paragraph by paragraph about AI use. Section-level attribution recorded at the point the expert finalises the draft — rather than reconstructed under cross — is the practical answer to a duty that now spans every forum your matter might land in.
CTA
The Expert Evidence AI Attribution Agent is in scoping. We’re working through pricing structure (per-report, per-expert subscription, or firm-licence) with design partners. Join the waitlist and we’ll let you know when access opens — your input on how Melbourne expert briefs are actually assembled will shape how the tier you sit in works.
Sources
- 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
- Administrative Review Tribunal — Practice Directions and Other Guidance for Professionals and Practitioners: https://www.art.gov.au/help-and-resources/professionals-and-practitioners/practice-directions-and-other-guidance
- Law Council of Australia — Australian Solicitors’ Conduct Rules: https://lawcouncil.au/policy-agenda/regulation-of-the-profession-and-ethics/australian-solicitors-conduct-rules
- Federal Court of Australia — Rules, Acts & Regulations: https://www.fedcourt.gov.au/law-and-practice/rules-acts-and-regulations
Exegesis capability references: