--- title: "Commercial Litigation & Dispute Resolution" url: "https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution" type: "practice_area" site: "AI Law Framework" id: "A5" task_count: 30 last_reviewed: "2026-10-10" --- # Commercial Litigation & Dispute Resolution > Organise evidence and preparation around the case theory, court requirements and client instructions.… Canonical page: https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution Last reviewed: 2026-10-10 Site index for agents: https://ailawframework.com/llms.txt Also covers: Commercial disputes, civil litigation, dispute resolution, ADR (mediation, arbitration, expert determination), class actions, insolvency litigation, QCAT/NCAT proceedings, Federal Court litigation, Supreme Court litigation Commercial litigation involves court-based and ADR proceedings to resolve business disputes, from pleadings through judgment or settlement; tasks span highly structured procedural steps (limitation deadlines, service, discovery) and open-ended strategic and advocacy work. The AI risk profile is among the most severe in legal practice: court filings carry professional conduct obligations enforced by real disciplinary outcomes (Re Dayal — first Australian practitioner sanctioned, August 2025: certificate varied, must work under supervision), hallucinated authorities are now addressed by mandatory practice notes in every major Australian jurisdiction, and privilege/confidentiality exposure in discovery is acute — the "policy theatre" failure mode is especially dangerous here because informal daily AI use (drafting affidavits, researching citations via ChatGPT) sits directly alongside the highest-consequence documents a lawyer produces. ## Key facts - Tasks: 30 - Green: 4, Yellow: 16, Red: 10 - Levels: Green is AI drafts and the lawyer checks; Yellow is lawyer and AI draft together; Red is the lawyer writes and AI may only check, clarify or collect data. ## Summary Organise evidence and preparation around the case theory, court requirements and client instructions. Check every relied-on authority and factual proposition, while separately looking for missing or contradictory material. ## Principles - **Choose the court before choosing the workflow**: Court AI rules differ. The NSW Supreme Court and Victorian Supreme Court publish separate directions, including different approaches to witness evidence and disclosure. Confirm the current direction and any case-specific order. - **Make every important point traceable**: A useful litigation output links issues, documents, authorities and quotations back to precise sources. A fluent summary without those links creates more review work. - **Separate preparation from evidence**: Document organisation and research tables can assist preparation. They do not establish admissibility, prove a fact or authorise generated witness evidence. ## Checklist - Open each relied-on authority independently. Confirm its existence, current status, proposition and any quotation or pinpoint reference. - Confirm each factual assertion with the witness and original record. Preserve the witness’s own knowledge and words, and check the court’s rules for this document. - Trace every relied-on figure to its source, confirm dates and assumptions, and independently recalculate totals or derived amounts. - Calculate operative dates independently from the current rule and actual triggering event. Check exceptions and record the verified deadline. - Check the current court or regulator requirements, prescribed form and attachments for the intended document and filing. - Review operative clauses in the complete document, including definitions, cross-references, schedules and inconsistencies with the instructions. - Confirm that the wording reflects the client’s instructions and contains no unintended admission, concession or commitment. Check the recipient and tone. - Confirm that the actual tool, feature and processing terms permit use of this information, including access, retention, training and disclosure conditions. - Check the specialist’s assumptions, qualifications and limits. Obtain the relevant expert or specialist review before relying on a technical conclusion. - Review the original source set for relevant material absent from the output, including contrary evidence, qualifications and unresolved gaps. - Verify the applicable jurisdiction and the version of each relevant law, rule or instrument in force at the material time. - Resolve exceptions and record the responsible lawyer’s approval of the exact output and intended use. ## Restrictions - Confirm firm, client and applicable court or professional restrictions before using any AI feature. ## Jurisdiction note Apply this guidance to the actual matter and jurisdiction. Confirm current requirements before relying on an example. ## Basis Suggested practical workflows informed by current primary and professional sources. Confirm each source’s jurisdiction and scope. ## Current requirements Choose the court before choosing the workflow. Court AI rules differ. The NSW Supreme Court and Victorian Supreme Court publish separate directions, including different approaches to witness evidence and disclosure. Confirm the current direction and any case-specific order. ## Current requirements Make every important point traceable. A useful litigation output links issues, documents, authorities and quotations back to precise sources. A fluent summary without those links creates more review work. ## Current requirements Separate preparation from evidence. Document organisation and research tables can assist preparation. They do not establish admissibility, prove a fact or authorise generated witness evidence. Also covers: Commercial disputes, civil litigation, dispute resolution, ADR (mediation, arbitration, expert determination), class actions, insolvency litigation, QCAT/NCAT proceedings, Federal Court litigation, Supreme Court litigation Commercial litigation involves court-based and ADR proceedings to resolve business disputes, from pleadings through judgment or settlement; tasks span highly structured procedural steps (limitation deadlines, service, discovery) and open-ended strategic and advocacy work. The AI risk profile is among the most severe in legal practice: court filings carry professional conduct obligations enforced by real disciplinary outcomes (Re Dayal — first Australian practitioner sanctioned, August 2025: certificate varied, must work under supervision), hallucinated authorities are now addressed by mandatory practice notes in every major Australian jurisdiction, and privilege/confidentiality exposure in discovery is acute — the "policy theatre" failure mode is especially dangerous here because informal daily AI use (drafting affidavits, researching citations via ChatGPT) sits directly alongside the highest-consequence documents a lawyer produces. ## Limits on AI assistance Red lines — never in this area ## Limits on AI assistance AI must never draft the substantive content of affidavits, witness statements, or character references. This prohibition is now statutory-equivalent in NSW (SC Gen 23, 3 February 2025), Federal Court (GPN-AI, April 2026), Victorian Supreme Court (SC Gen 25, May 2026), and NCAT (PD7, April 2025). AI must not alter, embellish, strengthen, dilute, or rephrase a witness's evidence in any written form. AI-generated case citations must never be filed or submitted to any court or tribunal, or included in advice to a client, without independent verification in a primary legal database (AustLII, LexisNexis AU, Westlaw AU) that the case exists, says exactly what is claimed, and is currently good law on the material point. The first Australian practitioner sanctioned for this failure (Re Dayal, VLSB August 2025) lost the right to operate an independent practice for at least two years. Legal-specific AI tools are not exempt — Re Dayal involved a specialist legal AI platform. Client confidential and privileged communications must never be uploaded to public or consumer-grade AI platforms (ChatGPT, Gemini, Copilot in personal/unmanaged mode). This breaches ASCR r 9.1 and may waive LPP (Evidence Act 1995 (Cth/NSW) s 122). Even enterprise-grade tools must be checked: verify the data processing agreement prohibits training on and retention of matter data. Documents produced under compulsion in litigation (subject to the Harman/implied undertaking: Hearne v Street (2008) 235 CLR 125) must not be uploaded to any AI tool that may train on, retain, or process inputs for a purpose outside the specific proceeding. Using discovered documents to train or operate a general AI model is a use outside the proceeding and constitutes breach of the undertaking. Court leave is required before any such use. AI litigation outcome predictions and win-probability scores must never be communicated to clients as reliable estimates — Australian judicial analytics data is methodologically insufficient, and presenting AI-generated prospects figures as reliable advice may breach ASCR r 9.2.2 (not misleading clients about prospects). AI must never be the sole mechanism for calculating limitation periods or procedural deadlines. Limitation period errors are unrecoverable at the matter level and create direct professional negligence exposure. A qualified lawyer must personally perform and record the calculation. AI must never generate an expert's opinion, be incorporated into the substance of an expert's report, or produce content attributed to the expert's independent judgment. The expert must independently hold and express every opinion in any filed report. Where an expert uses AI to draft report content, prior leave of the court or tribunal is required (see NCAT PD7 and GPN-AI). AI must not be used to coach witnesses — providing AI-generated scripts of expected questions and answers for witnesses to memorise risks breach of ASCR rr 17 and 24.1 prohibiting coaching. The distinction between legitimate witness preparation and improper coaching is narrow; AI-generated question-and-answer frameworks cross that line. AI must not calculate or verify stamp duty, PPSA discharge requirements, or other regulatory completion steps in settlement documentation without independent legal verification — AI frequently omits jurisdiction-specific completion requirements in commercial settlement deeds. - AI must never draft the substantive content of affidavits, witness statements, or character references. This prohibition is now statutory-equivalent in NSW (SC Gen 23, 3 February 2025), Federal Court (GPN-AI, April 2026), Victorian Supreme Court (SC Gen 25, May 2026), and NCAT (PD7, April 2025). AI must not alter, embellish, strengthen, dilute, or rephrase a witness's evidence in any written form. - AI-generated case citations must never be filed or submitted to any court or tribunal, or included in advice to a client, without independent verification in a primary legal database (AustLII, LexisNexis AU, Westlaw AU) that the case exists, says exactly what is claimed, and is currently good law on the material point. The first Australian practitioner sanctioned for this failure (Re Dayal, VLSB August 2025) lost the right to operate an independent practice for at least two years. Legal-specific AI tools are not exempt — Re Dayal involved a specialist legal AI platform. - Client confidential and privileged communications must never be uploaded to public or consumer-grade AI platforms (ChatGPT, Gemini, Copilot in personal/unmanaged mode). This breaches ASCR r 9.1 and may waive LPP (Evidence Act 1995 (Cth/NSW) s 122). Even enterprise-grade tools must be checked: verify the data processing agreement prohibits training on and retention of matter data. - Documents produced under compulsion in litigation (subject to the Harman/implied undertaking: Hearne v Street (2008) 235 CLR 125) must not be uploaded to any AI tool that may train on, retain, or process inputs for a purpose outside the specific proceeding. Using discovered documents to train or operate a general AI model is a use outside the proceeding and constitutes breach of the undertaking. Court leave is required before any such use. - AI litigation outcome predictions and win-probability scores must never be communicated to clients as reliable estimates — Australian judicial analytics data is methodologically insufficient, and presenting AI-generated prospects figures as reliable advice may breach ASCR r 9.2.2 (not misleading clients about prospects). - AI must never be the sole mechanism for calculating limitation periods or procedural deadlines. Limitation period errors are unrecoverable at the matter level and create direct professional negligence exposure. A qualified lawyer must personally perform and record the calculation. - AI must never generate an expert's opinion, be incorporated into the substance of an expert's report, or produce content attributed to the expert's independent judgment. The expert must independently hold and express every opinion in any filed report. Where an expert uses AI to draft report content, prior leave of the court or tribunal is required (see NCAT PD7 and GPN-AI). - AI must not be used to coach witnesses — providing AI-generated scripts of expected questions and answers for witnesses to memorise risks breach of ASCR rr 17 and 24.1 prohibiting coaching. The distinction between legitimate witness preparation and improper coaching is narrow; AI-generated question-and-answer frameworks cross that line. - AI must not calculate or verify stamp duty, PPSA discharge requirements, or other regulatory completion steps in settlement documentation without independent legal verification — AI frequently omits jurisdiction-specific completion requirements in commercial settlement deeds. ## Jurisdiction-specific considerations Australian traps ## Jurisdiction-specific considerations Federal Court GPN-AI (effective 16 April 2026): Mandatory disclosure where AI has been used to generate, summarise, or analyse material informing evidence or submissions — disclosure must appear at the start of the document, naming the AI used and how. Queensland courts (Supreme Court PD 5 of 2025, District Court PD 12 of 2025): individual named practitioner (not just the firm) must be identified at the foot of all written submissions — a named-accountability mechanism targeting AI-anonymised documents. Non-compliance across all jurisdictions may result in adverse costs orders and professional disciplinary referral. NSW Supreme Court Practice Note SC Gen 23 (effective 3 February 2025) and Victorian Supreme Court SC Gen 25 (effective 14 May 2026): Explicit prohibition on AI generating the substance of affidavits, witness statements, and character references. Both require a disclosure on each such document confirming AI was not used to generate its content. NCAT Procedural Direction PD7 (April 2025) mirrors SC Gen 23 for tribunal proceedings. NSW Personal Injury Commission PIC13 applies equivalent requirements to that forum. Check the current practice note for every court/tribunal before filing any document. Re Dayal (VLSB, August 2025 — first Australian sanction for AI hallucination): Practitioner submitted AI-generated false citations to the Federal Circuit and Family Court; referred by Justice Humphreys to VLSB+C; outcome: practicing certificate varied — practitioner can no longer operate as a principal, manage trust accounts, or practise independently; must work under supervision for at least two years. The VLSB found that ignorance of an AI tool's hallucination risk is not a defence to the competence obligation. This case applies equally to specialist legal AI tools, not only to general consumer LLMs. Limitation period complexity — no single Australian statute: Each state and territory has separate Acts with different periods, accrual rules, and extension provisions. Periods relevant to commercial litigation: 6 years (contract, most jurisdictions); 12 years (deeds, NSW/Vic — note Queensland reduced to 6 years under Property Law Act 2023 (Qld) from August 2025); 6 years (statutory debt, Corporations Act claims); 3 years (personal injury, most jurisdictions); 2 years (contribution claims, Law Reform (Miscellaneous Provisions) Act equivalents). AI models are unreliable on jurisdiction-specific and recently amended limitation rules — the Queensland August 2025 changes are a verified AI failure point. LPP waiver through AI tool use: Federal Court GPN-AI and Victorian SC Gen 25 explicitly flag risk of inadvertent LPP waiver where privileged material is uploaded to AI tools even described as 'enterprise' or 'closed'. Early US decisions (United States v Heppner, February 2026) refused privilege over AI-generated documents where the AI platform's data-collection terms meant confidentiality could not be established. Australian courts have not yet decided the point but the trajectory is clear. Before any upload of privileged material to any AI tool: verify the vendor's data processing agreement prohibits training on and retention of inputs, review access controls, and document the verification. Harman/implied undertaking (Hearne v Street (2008) 235 CLR 125): Documents produced under compulsion in litigation may only be used for the purposes of that proceeding. Uploading discovered documents to an external AI tool that retains or processes data outside the matter is a use for an extraneous purpose and a breach of the undertaking — contempt exposure and adverse costs. This applies to all AI-assisted discovery tools operating on external cloud infrastructure unless the contract explicitly confines data use to the specific matter and the vendor has no training or retention rights. ASCR professional conduct obligations: r 4.1.3 (competence — Law Council and state/territory regulators have confirmed technological literacy, including understanding AI limitations, is now part of the competence duty; Re Dayal confirms ignorance is no defence); r 9.1 (confidentiality — uploading client information to public AI is likely a breach); r 17 (duties to court — filing AI-hallucinated citations risks a finding of misleading conduct); r 19.1 (frankness and candour — a practitioner cannot file submissions containing authorities they have not individually verified); r 24.1 (not coaching witnesses — AI-generated witness preparation scripts risk this breach). Court-by-court disclosure position as at June 2026: Federal Court — GPN-AI mandatory disclosure for AI-informed evidence and submissions; NSW Supreme Court — SC Gen 23 mandatory AI non-use disclosure on affidavits/witness statements; NSW NCAT — PD7 mandatory AI non-use disclosure; Victorian Supreme Court — SC Gen 25 mandatory AI non-use disclosure on sworn documents, general verification obligation; Queensland Supreme Court — PD 5 of 2025 (named practitioner at foot of submissions); Queensland District Court — PD 12 of 2025 (same); Queensland Civil and Administrative Tribunal (QCAT) — AI hallucination guidelines in force but no equivalent practitioner-facing prohibition as at June 2026. No single national rule — check the current practice note for the specific court before filing. Costs and billing ethics: College of Law guidance (2025) and ASCR r 4.1.2 require that time savings from AI be reflected in fees charged — billing as though the task required human time when AI produced it in a fraction is a potential costs-assessment ground for reduction and a professional conduct risk. Cost assessors in Supreme Courts and the Federal Court are beginning to scrutinise AI-assisted time entries. Firms must have a documented AI billing policy before using AI in billable matters. AI and arbitration — no specific Australian rules as at June 2026 but emerging: ACICA is reviewing its 2021 Rules and the International Arbitration Conference (ACICA, October 2025) specifically addressed AI evidence fabrication risks. Chief Justice Bell (NSW) called on arbitrators and counsel to address enhanced risks of AI-generated false evidence. ACICA rules currently do not address AI disclosure — practitioners should consider whether the arbitration agreement or procedural orders should fill this gap, and apply the same verification disciplines as for court proceedings. - Federal Court GPN-AI (effective 16 April 2026): Mandatory disclosure where AI has been used to generate, summarise, or analyse material informing evidence or submissions — disclosure must appear at the start of the document, naming the AI used and how. Queensland courts (Supreme Court PD 5 of 2025, District Court PD 12 of 2025): individual named practitioner (not just the firm) must be identified at the foot of all written submissions — a named-accountability mechanism targeting AI-anonymised documents. Non-compliance across all jurisdictions may result in adverse costs orders and professional disciplinary referral. - NSW Supreme Court Practice Note SC Gen 23 (effective 3 February 2025) and Victorian Supreme Court SC Gen 25 (effective 14 May 2026): Explicit prohibition on AI generating the substance of affidavits, witness statements, and character references. Both require a disclosure on each such document confirming AI was not used to generate its content. NCAT Procedural Direction PD7 (April 2025) mirrors SC Gen 23 for tribunal proceedings. NSW Personal Injury Commission PIC13 applies equivalent requirements to that forum. Check the current practice note for every court/tribunal before filing any document. - Re Dayal (VLSB, August 2025 — first Australian sanction for AI hallucination): Practitioner submitted AI-generated false citations to the Federal Circuit and Family Court; referred by Justice Humphreys to VLSB+C; outcome: practicing certificate varied — practitioner can no longer operate as a principal, manage trust accounts, or practise independently; must work under supervision for at least two years. The VLSB found that ignorance of an AI tool's hallucination risk is not a defence to the competence obligation. This case applies equally to specialist legal AI tools, not only to general consumer LLMs. - Limitation period complexity — no single Australian statute: Each state and territory has separate Acts with different periods, accrual rules, and extension provisions. Periods relevant to commercial litigation: 6 years (contract, most jurisdictions); 12 years (deeds, NSW/Vic — note Queensland reduced to 6 years under Property Law Act 2023 (Qld) from August 2025); 6 years (statutory debt, Corporations Act claims); 3 years (personal injury, most jurisdictions); 2 years (contribution claims, Law Reform (Miscellaneous Provisions) Act equivalents). AI models are unreliable on jurisdiction-specific and recently amended limitation rules — the Queensland August 2025 changes are a verified AI failure point. - LPP waiver through AI tool use: Federal Court GPN-AI and Victorian SC Gen 25 explicitly flag risk of inadvertent LPP waiver where privileged material is uploaded to AI tools even described as 'enterprise' or 'closed'. Early US decisions (United States v Heppner, February 2026) refused privilege over AI-generated documents where the AI platform's data-collection terms meant confidentiality could not be established. Australian courts have not yet decided the point but the trajectory is clear. Before any upload of privileged material to any AI tool: verify the vendor's data processing agreement prohibits training on and retention of inputs, review access controls, and document the verification. - Harman/implied undertaking (Hearne v Street (2008) 235 CLR 125): Documents produced under compulsion in litigation may only be used for the purposes of that proceeding. Uploading discovered documents to an external AI tool that retains or processes data outside the matter is a use for an extraneous purpose and a breach of the undertaking — contempt exposure and adverse costs. This applies to all AI-assisted discovery tools operating on external cloud infrastructure unless the contract explicitly confines data use to the specific matter and the vendor has no training or retention rights. - ASCR professional conduct obligations: r 4.1.3 (competence — Law Council and state/territory regulators have confirmed technological literacy, including understanding AI limitations, is now part of the competence duty; Re Dayal confirms ignorance is no defence); r 9.1 (confidentiality — uploading client information to public AI is likely a breach); r 17 (duties to court — filing AI-hallucinated citations risks a finding of misleading conduct); r 19.1 (frankness and candour — a practitioner cannot file submissions containing authorities they have not individually verified); r 24.1 (not coaching witnesses — AI-generated witness preparation scripts risk this breach). - Court-by-court disclosure position as at June 2026: Federal Court — GPN-AI mandatory disclosure for AI-informed evidence and submissions; NSW Supreme Court — SC Gen 23 mandatory AI non-use disclosure on affidavits/witness statements; NSW NCAT — PD7 mandatory AI non-use disclosure; Victorian Supreme Court — SC Gen 25 mandatory AI non-use disclosure on sworn documents, general verification obligation; Queensland Supreme Court — PD 5 of 2025 (named practitioner at foot of submissions); Queensland District Court — PD 12 of 2025 (same); Queensland Civil and Administrative Tribunal (QCAT) — AI hallucination guidelines in force but no equivalent practitioner-facing prohibition as at June 2026. No single national rule — check the current practice note for the specific court before filing. - Costs and billing ethics: College of Law guidance (2025) and ASCR r 4.1.2 require that time savings from AI be reflected in fees charged — billing as though the task required human time when AI produced it in a fraction is a potential costs-assessment ground for reduction and a professional conduct risk. Cost assessors in Supreme Courts and the Federal Court are beginning to scrutinise AI-assisted time entries. Firms must have a documented AI billing policy before using AI in billable matters. - AI and arbitration — no specific Australian rules as at June 2026 but emerging: ACICA is reviewing its 2021 Rules and the International Arbitration Conference (ACICA, October 2025) specifically addressed AI evidence fabrication risks. Chief Justice Bell (NSW) called on arbitrators and counsel to address enhanced risks of AI-generated false evidence. ACICA rules currently do not address AI disclosure — practitioners should consider whether the arbitration agreement or procedural orders should fill this gap, and apply the same verification disciplines as for court proceedings. ## Risks to address How it goes wrong ## Risks to address Hallucinated case citations filed with a court: AI generates plausible-sounding but non-existent Australian citations, or real cases with invented holdings or quotes. Re Dayal (VLSB, August 2025) established that this is sanctionable even where the AI tool was a specialist legal product. Consequences: indemnity costs orders, referral to VLSB+C/OLSC/equivalent, practice certificate conditions, reputational damage. Over 45 documented Australian instances (2023–2026). Stale or wrong-jurisdiction legislation: AI cites a superseded version of an Act, an interstate equivalent, or a provision not yet commenced. High failure risk on recently amended limitation periods (e.g. Queensland Property Law Act 2023 limitation changes, August 2025) and on jurisdiction-specific versions of Consumer Law, corporate insolvency provisions, and state-based civil procedure rules. LPP waiver through AI tool use: Lawyer uploads client instructions, draft pleadings, or litigation strategy to a public AI platform; the provider's data-collection terms undermine confidentiality; privilege is later contested on the basis that information was voluntarily disclosed to a third party. Early US precedents (United States v Heppner, February 2026) have refused privilege in analogous circumstances; Australian courts have not yet decided the point but the risk trajectory is clearly established. Harman undertaking breach: Discovered documents uploaded to an external AI tool for summarisation or review; tool retains or processes data outside the matter; constitutes use of discovered material for an extraneous purpose in breach of the implied undertaking (Hearne v Street). Contempt exposure and adverse costs. Missed limitation period: AI provides an incorrect expiry date (wrong Limitation Act, wrong cause of action period, wrong accrual date, or missed discoverability extension). Limitation-barred claim filed; professional negligence action against the solicitor; unrecoverable at the matter level. This is the highest-consequence AI failure mode in litigation. AI-drafted affidavit content tendered as witness evidence: Paralegal or junior uses AI to draft the substance of a witness statement; deponent signs without reading carefully; document contains AI-fabricated facts, reconstructed dialogue, or embellished events not in the deponent's actual recollection. Cross-examination destroys credibility; costs orders follow; potential disciplinary referral for breaching SC Gen 23/GPN-AI/SC Gen 25. - Hallucinated case citations filed with a court: AI generates plausible-sounding but non-existent Australian citations, or real cases with invented holdings or quotes. Re Dayal (VLSB, August 2025) established that this is sanctionable even where the AI tool was a specialist legal product. Consequences: indemnity costs orders, referral to VLSB+C/OLSC/equivalent, practice certificate conditions, reputational damage. Over 45 documented Australian instances (2023–2026). - Stale or wrong-jurisdiction legislation: AI cites a superseded version of an Act, an interstate equivalent, or a provision not yet commenced. High failure risk on recently amended limitation periods (e.g. Queensland Property Law Act 2023 limitation changes, August 2025) and on jurisdiction-specific versions of Consumer Law, corporate insolvency provisions, and state-based civil procedure rules. - LPP waiver through AI tool use: Lawyer uploads client instructions, draft pleadings, or litigation strategy to a public AI platform; the provider's data-collection terms undermine confidentiality; privilege is later contested on the basis that information was voluntarily disclosed to a third party. Early US precedents (United States v Heppner, February 2026) have refused privilege in analogous circumstances; Australian courts have not yet decided the point but the risk trajectory is clearly established. - Harman undertaking breach: Discovered documents uploaded to an external AI tool for summarisation or review; tool retains or processes data outside the matter; constitutes use of discovered material for an extraneous purpose in breach of the implied undertaking (Hearne v Street). Contempt exposure and adverse costs. - Missed limitation period: AI provides an incorrect expiry date (wrong Limitation Act, wrong cause of action period, wrong accrual date, or missed discoverability extension). Limitation-barred claim filed; professional negligence action against the solicitor; unrecoverable at the matter level. This is the highest-consequence AI failure mode in litigation. - AI-drafted affidavit content tendered as witness evidence: Paralegal or junior uses AI to draft the substance of a witness statement; deponent signs without reading carefully; document contains AI-fabricated facts, reconstructed dialogue, or embellished events not in the deponent's actual recollection. Cross-examination destroys credibility; costs orders follow; potential disciplinary referral for breaching SC Gen 23/GPN-AI/SC Gen 25. ## Review checklist The lawyer's review checklist ## Review checklist Citation integrity: Every case name, citation, principle, and pinpoint quote in any AI-assisted document has been individually confirmed in AustLII, LexisNexis AU, or Westlaw AU — the case exists, is reported correctly, says what is attributed to it, and has not been overruled or distinguished on the material point. No AI tool's own confirmation of a citation is sufficient. Affidavit/witness statement AI non-use compliance: No substantive content of any affidavit, witness statement, or character reference was generated or rephrased by AI; all content reflects the deponent's own knowledge and words; the document includes the court-required AI non-use disclosure where applicable (NSW SC Gen 23; Vic SC Gen 25; NCAT PD7; Federal Court GPN-AI); AI non-use disclosure is present on the document before filing. Privilege and confidentiality firewall: No client-confidential, legally privileged, or court-suppressed material was uploaded to any AI platform without first verifying that the vendor's data processing agreement prohibits training on and external retention of inputs; confirm whether an enterprise contract (not just a consumer-mode setting) is in force; document this check in the file. Harman undertaking compliance: No documents produced under compulsion in this or any other proceeding were fed into any AI tool in a manner that constitutes use outside the proceeding; where any discovered documents were used as AI input, confirm the contract confines data use to this matter and no training or retention rights apply. Limitation periods and court deadlines independently calculated: The relevant limitation period (and any interlocutory deadlines) has been calculated by the lawyer personally, by reference to the correct state or territory Limitation Act, the specific cause of action, the accrual date, and any applicable discoverability extension, disability suspension, or acknowledgment restart provision; this calculation has been recorded in the file and diarised with advance alerts. Court-specific AI disclosure compliance: The document complies with the current practice note for the specific court and document type — GPN-AI disclosure for Federal Court evidence and submissions; SC Gen 23 disclosure for NSW Supreme Court affidavits; SC Gen 25 disclosure for Victorian Supreme Court sworn documents; named practitioner at foot of submissions for Queensland Supreme and District Courts; PD7 for NCAT; check equivalent for any other court or tribunal before filing. Client instructions confirmed on all material matters: All facts, settlement parameters, risk-tolerance positions, and strategy decisions in any AI-drafted output have been confirmed in writing with the client; AI has not assumed instructions the client has not given; the advice letter on prospects is the lawyer's own professional judgment, not an AI-generated assessment. Expert opinion independence verified: Any AI involvement in expert-related documents was limited to structuring questions or preparing the briefing letter (not its expert-opinion content); the expert's stated opinion is entirely their own; the expert declaration/certificate complies with the applicable code; any AI use by the expert was disclosed and (where required) leave was obtained. - Citation integrity: Every case name, citation, principle, and pinpoint quote in any AI-assisted document has been individually confirmed in AustLII, LexisNexis AU, or Westlaw AU — the case exists, is reported correctly, says what is attributed to it, and has not been overruled or distinguished on the material point. No AI tool's own confirmation of a citation is sufficient. - Affidavit/witness statement AI non-use compliance: No substantive content of any affidavit, witness statement, or character reference was generated or rephrased by AI; all content reflects the deponent's own knowledge and words; the document includes the court-required AI non-use disclosure where applicable (NSW SC Gen 23; Vic SC Gen 25; NCAT PD7; Federal Court GPN-AI); AI non-use disclosure is present on the document before filing. - Privilege and confidentiality firewall: No client-confidential, legally privileged, or court-suppressed material was uploaded to any AI platform without first verifying that the vendor's data processing agreement prohibits training on and external retention of inputs; confirm whether an enterprise contract (not just a consumer-mode setting) is in force; document this check in the file. - Harman undertaking compliance: No documents produced under compulsion in this or any other proceeding were fed into any AI tool in a manner that constitutes use outside the proceeding; where any discovered documents were used as AI input, confirm the contract confines data use to this matter and no training or retention rights apply. - Limitation periods and court deadlines independently calculated: The relevant limitation period (and any interlocutory deadlines) has been calculated by the lawyer personally, by reference to the correct state or territory Limitation Act, the specific cause of action, the accrual date, and any applicable discoverability extension, disability suspension, or acknowledgment restart provision; this calculation has been recorded in the file and diarised with advance alerts. - Court-specific AI disclosure compliance: The document complies with the current practice note for the specific court and document type — GPN-AI disclosure for Federal Court evidence and submissions; SC Gen 23 disclosure for NSW Supreme Court affidavits; SC Gen 25 disclosure for Victorian Supreme Court sworn documents; named practitioner at foot of submissions for Queensland Supreme and District Courts; PD7 for NCAT; check equivalent for any other court or tribunal before filing. - Client instructions confirmed on all material matters: All facts, settlement parameters, risk-tolerance positions, and strategy decisions in any AI-drafted output have been confirmed in writing with the client; AI has not assumed instructions the client has not given; the advice letter on prospects is the lawyer's own professional judgment, not an AI-generated assessment. - Expert opinion independence verified: Any AI involvement in expert-related documents was limited to structuring questions or preparing the briefing letter (not its expert-opinion content); the expert's stated opinion is entirely their own; the expert declaration/certificate complies with the applicable code; any AI use by the expert was disclosed and (where required) leave was obtained. ## Tasks (30) | Task | Level | How AI is used | What the lawyer must review | | --- | --- | --- | --- | | [Build an issue-to-document evidence table](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/build-an-issue-to-document-evidence-table.md) | Yellow | Organise permitted documents against issues defined by the lawyer, retaining exact page or paragraph references and competing material. | Check the complete relevant record Decide what the evidence supports | | [Prepare a navigable hearing bundle index](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/prepare-a-navigable-hearing-bundle-index.md) | Yellow | Draft a document index from a final lawyer-selected set, identify apparent duplicates and check proposed page links. | Check every final index entry Confirm the bundle meets the applicable court order or practice requirements | | [Check authorities and evidence before filing submissions](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/check-authorities-and-evidence-before-filing-submissions.md) | Yellow | Extract citations, quotations and evidence references from a draft into a verification table. | Verify each reference directly in reliable sources Record the required statement in the submissions where the court requires one | | [Convert court orders into an action and responsibility list](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/convert-court-orders-into-an-action-and-responsibility-list.md) | Red | Tabulate each paragraph of the sealed orders and every later variation into the internal action list, carrying across the step ordered, the party the order… | The lawyer reads the sealed orders and every variation, settles each row of the action list and the responsible person on it, and traces each row back to its… | | [Draft pleadings from the facts-and-issues brief](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-pleadings-from-the-facts-and-issues-brief.md) | Yellow | From the facts-and-issues brief, draft a pleading or defence with material facts mapped to each element and a list of particulars still needed. | Confirm every pleaded fact has instructions or evidence on the file Check the court's pleading rules, limitation and the causes of action before filing | | [Draft an affidavit or witness statement from the proof](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-an-affidavit-or-witness-statement-from-the-proof.md) | Red | The tool drafts no part of the affidavit or witness statement and suggests no wording for it, in the document or beside it: no skeleton, no structure, no… | The lawyer drafts the evidence in the witness's own words from the proof, and the witness must read and adopt it before it is sworn or signed.… | | [Review the discovery set and draft a production issues list](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/review-the-discovery-set-and-draft-a-production-issues-list.md) | Yellow | From the permitted discovery set, group documents, flag likely privileged or sensitive items not yet ruled on, and draft a first-pass relevance and issues list. | Make privilege and production calls yourself Do not load discovered or suppressed material into a tool without an express decision | | [Draft written submissions from the evidence table](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-written-submissions-from-the-evidence-table.md) | Yellow | Build a structured outline of argument with proposed record citations from the evidence table and authorities list the lawyer supplied. | Verify every citation in a trusted source and against the transcript or exhibit Do not file AI-proposed authorities you have not read | | [Draft a letter of demand or without-prejudice settlement offer](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-a-letter-of-demand-or-without-prejudice-settlement-offer.md) | Yellow | Draft a first-pass letter of demand or without-prejudice settlement offer from the approved issues table, with a list of figures that still need a source. | Get written client authority for any offer Check form, service and costs-consequence rules for the court before anything is sent | | [Draft an interlocutory application, supporting evidence and outline](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-an-interlocutory-application-supporting-evidence-and-outline.md) | Yellow | Draft the interlocutory process and a submissions outline from the relief sought and the facts brief.… | Confirm the correct process for that court and the urgency Leave the supporting affidavit to the deponent Disclose adverse facts yourself on without-notice… | | [Prepare a hearing chronology from the key documents](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/prepare-a-hearing-chronology-from-the-key-documents.md) | Yellow | Build a dated chronology with document references from the permitted pleadings and key files, and list gaps and inconsistencies. | Source every row on the file Leave out argument unless the court asks for a narrative chronology | | Draft the expert brief and review the draft report for scope | Yellow | Turn the issues list into a draft letter of instruction and, once a draft report arrives, table assumptions, opinions and any AI-disclosure language to check. | Confirm instructions and assumptions Confirm the report answers the questions asked Remember the expert's duty is to the court | | Draft the mediation position paper and a settlement-range briefing note | Yellow | From the pleadings, evidence table and instructions, draft a mediation paper and a confidential BATNA/WATNA note for the lawyer. | Take settlement range and offers from client instructions Check the paper does not waive privilege or overstate the evidence | | [Draft the settlement deed, consent orders and releases](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-the-settlement-deed-consent-orders-and-releases.md) | Yellow | Turn the settlement terms into a first-draft deed, consent orders and release clauses, with a list of parties, proceeding details and completion steps to… | Check legal names, execution formalities and any court-approval requirement Check release scope;… | | Build a limitation and procedural-deadline calendar from the file | Red | Extract filing, service, limitation and timetable dates from the pleadings, orders and correspondence into a calendar with source references. | Calculate every deadline from the current court rules and the sealed orders Do not treat date arithmetic as authority to miss a limit | | [Check a court document against the AI practice note before filing](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/check-a-court-document-against-the-ai-practice-note-before-filing.md) | Red | Extract citations, quotes and chronology rows from a draft court document into a verification list. | Check the draft against the practice note of the court on the file Verify every citation in a source you trust | | [Decide whether discovery or suppressed material can go into a tool](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/decide-whether-discovery-or-suppressed-material-can-go-into-a-tool.md) | Red | Build an index of candidate documents from metadata the lawyer has already cleared, without opening the underlying files in the tool. | Decide whether discovery, subpoenaed or suppressed material may be processed, and where Do not treat the index as permission to upload | | [Check legislation citations against the official register](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/check-legislation-citations-against-the-official-register.md) | Red | Extract legislation citations from a draft into a list for checking on the official register. | Open each citation on this jurisdiction's official legislation register Do not treat a second model as confirmation | | [Check procedural citations against the court rules](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/check-procedural-citations-against-the-court-rules.md) | Red | Extract procedural citations from a draft into a list for checking against the court rules. | Open the court rules that apply in this jurisdiction Do not paste interstate procedure as if it were local | | Summarise the opposing pleading and list the facts still in issue | Green | Turn the claim and defence into a table of admitted, denied and not-admitted facts, with a list of particulars that remain outstanding. | Check the table against the filed pleadings Decide the legal effect of admissions; do not treat a summary as the pleading | | [Check an NCAT filing for material barred from Gen AI tools](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/check-an-ncat-filing-for-material-barred-from-gen-ai-tools.md) | Red | List summonsed, non-publication and statutorily restricted material in the file. | Decide whether any controlled-environment exception applies (para 9). | | Prepare an SA court affidavit compliant with the SA Gen AI guidelines | Red | Check that any AI-assisted statement fairly and fully reflects the witness's evidence. | Confirm with the witness. | | [Verify every authority in a brief before filing](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/verify-every-authority-in-a-brief-before-filing.md) | Red | Check AI-produced citations against authorised reports. | Costs and referrals follow fake citations. | | [Prepare an expert-report leave application where Gen AI will be used](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/prepare-an-expert-report-leave-application-where-gen-ai-will-be-used.md) | Yellow | Draft the application for leave to use Gen AI in an expert report. | NSW SC Gen 23 para 20 requires leave. | | [Investigate AI tools for document collation as ordered by the court](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/investigate-ai-tools-for-document-collation-as-ordered-by-the-court.md) | Yellow | Assess AI for chronologies, summaries and document management. | Follow court orders like Lee J's s 37P(2) orders. | | [Draft a firm AI-supervision protocol for litigation teams](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/draft-a-firm-ai-supervision-protocol-for-litigation-teams.md) | Green | Draft supervision and verification rules. | Partner approves. | | [First-pass document review for relevance in discovery](https://ailawframework.com/practice/commercial-litigation-and-dispute-resolution/first-pass-document-review-for-relevance-in-discovery.md) | Yellow | Use AI to rank and classify documents. | Lawyer reviews privileged and borderline documents. | | Prepare an AI-use disclosure statement for a Federal Court filing | Green | Draft disclosure per GPN-AI. | Lawyer confirms compliance. | | Verify every authority in submissions before filing | Green | Run a citation check against AustLII/JADE. | Lawyer signs off; see the AI hallucination cases. | | Summarise discovery for privilege review with technology-assisted review | Yellow | Run TAR on a discovery set. | Lawyer reviews privilege calls. | ## Jurisdiction notes ### Australia #### Witness evidence AI must not draft the substance of affidavits, witness statements or character references, or alter a witness's evidence. NSW SC Gen 23 (3 February 2025), GPN-AI (April 2026), Vic SC Gen 25 (May 2026), NCAT PD7 (April 2025). #### Citations Check every case on AustLII, LexisNexis AU or Westlaw AU: it exists, says what is claimed and is good law. Re Dayal (VLSB, August 2025): no independent practice for at least two years; a legal AI tool was used. #### Experts Every opinion in a filed report must be the expert's own. An expert who uses AI to draft report content needs prior leave (NCAT PD7, GPN-AI). - Never use AI to draft, embellish or rephrase the substance of a witness's evidence. - Never file, submit or advise on an AI-generated citation without checking it in a primary database. - Never let AI generate an expert's opinion or the substance of an expert's report. Sources: - [Supreme Court of NSW: SC Gen 23, amended 28 January 2025](https://supremecourt.nsw.gov.au/documents/Practice-and-Procedure/Practice-Notes/general/current/PN_SC_Gen_23.pdf) (checked 2026-09-16) - [Supreme Court of Victoria: SC GEN 25, 14 May 2026](https://www.supremecourt.vic.gov.au/areas/legal-resources/practice-notes/sc-gen-25-the-use-of-artificial-intelligence-by-court-users) (checked 2026-09-16) ### New Zealand #### Court AI guidelines Open the Courts of New Zealand lawyer guidelines (7 December 2023) before filing. They cover the Supreme Court, Court of Appeal, High Court, District Court and listed specialist courts and tribunals. #### Known risk Chatbots appear to have had limited training on New Zealand law and procedure. They can invent cases and then confirm the invention. #### Your duties Check every citation before use. You do not have to disclose GenAI use unless the court asks. #### Fake cases The Law Society recorded in December 2025 that New Zealand courts are already seeing references to cases that do not exist. #### Procedure and checking High Court Rules 2016 for civil procedure. Check legislation on legislation.govt.nz and cases in a trusted database or via the Law Society Library. Do not apply SC Gen 23 or GPN-AI. - Do not file a generated citation, quote or statutory reference until you have opened it in legislation.govt.nz or a trusted case database. - Do not enter private, confidential, suppressed or privileged material into a public chatbot. - Do not alter a witness's evidence with a generative tool. Sources: - [Courts of New Zealand: GenAI guidelines — Lawyers (7 December 2023)](https://www.courtsofnz.govt.nz/assets/6-Going-to-Court/practice-directions/practice-guidelines/all-benches/20231207-GenAI-Guidelines-Lawyers.pdf) (checked 2026-09-23) - [Courts of New Zealand: Guidelines for use of generative AI in courts and tribunals](https://www.courtsofnz.govt.nz/going-to-court/practice-directions/practice-guidelines/all-benches/guidelines-for-use-of-generative-artificial-intelligence-in-courts-and-tribunals) (checked 2026-09-23) - [High Court Rules 2016](https://www.legislation.govt.nz/regulation/public/2016/0225/104.0/whole.html) (checked 2026-09-23) - [NZLS: Generative AI guidance for lawyers (March 2024)](https://www.lawsociety.org.nz/professional-practice/rules-and-maintaining-professional-standards/generative-ai-guidance-for-lawyers/) (checked 2026-09-23) - [NZLS: Lawyers are responsible for the contents of court documents prepared using generative AI](https://www.lawsociety.org.nz/news/newsroom/lawyers-are-responsible-for-the-contents-of-court-documents-prepared-using-generative-ai/) (checked 2026-09-23) - [NZLS: AI Hub](https://www.lawsociety.org.nz/professional-practice/aihub/) (checked 2026-09-23) ### United States Local name: Commercial litigation and dispute resolution #### Rule 11 Federal civil filings certify, after reasonable inquiry, that legal contentions are warranted by existing law. Mata v. Avianca sanctioned counsel for ChatGPT-fabricated authorities. #### Standing orders Some judges add one. An E.D. Tex. order of 9 April 2025 requires a generative-AI certificate; a filing without it is presumed to use no generative AI. #### Discovery Scope is Rule 26(b)(1) proportionality, including ESI. Mental impressions are protected work product (Rule 26(b)(3)(B)); D.C. Opinion 388 warns against public tools. #### Before you file Open the judge's standing orders. Verify every authority in a reporter or official database, not with another generative tool. - Do not file a generated citation you have not opened in a real reporter. - Do not put privileged discovery or work product into a public tool. Sources: - [Federal Rule of Civil Procedure 11](https://www.law.cornell.edu/rules/frcp/rule_11) (checked 2026-09-23) - [Federal Rule of Civil Procedure 26](https://www.law.cornell.edu/rules/frcp/rule_26) (checked 2026-09-23) - [Mata v. Avianca, Inc., Opinion and Order on Sanctions (S.D.N.Y. 22 June 2023)](https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2022cv01461/575368/54/) (checked 2026-09-23) - [E.D. Tex., Standing Order on generative artificial intelligence (9 April 2025)](https://coop.txed.uscourts.gov/sites/default/files/judgeFiles/JDL%20Standing%20Order%20on%20AI%204.9.25.pdf) (checked 2026-09-23) - [District of Columbia Bar Ethics Opinion 388](https://www.dcbar.org/for-lawyers/legal-ethics/ethics-opinions-210-present/ethics-opinion-388) (checked 2026-09-23) - [ABA Formal Opinion 512, Generative Artificial Intelligence Tools (29 July 2024)](https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf) (checked 2026-09-23) ## Sources - [Supreme Court of NSW: SC Gen 23, amended 28 January 2025](https://supremecourt.nsw.gov.au/documents/Practice-and-Procedure/Practice-Notes/general/current/PN_SC_Gen_23.pdf) (checked 2026-09-16) - [Supreme Court of Victoria: SC GEN 25, 14 May 2026](https://www.supremecourt.vic.gov.au/areas/legal-resources/practice-notes/sc-gen-25-the-use-of-artificial-intelligence-by-court-users) (checked 2026-09-16) *Practice guidance for planning and review. Confirm current primary sources and the requirements of the actual matter. This is general information, not legal advice.*