MFRD and National Disability Insurance Agency (NDIS), [2026] ARTA 1833
·Administrative Review Tribunal of Australia·AUWarning
In Australia's Administrative Review Tribunal, a self-represented NDIS applicant filed submissions prepared with generative AI that contained fictitious tribunal decisions and inaccurate statutory references. General Member Strange warned that false AI-generated material can hamper tribunal decision-making, then decided the merits and remitted the support decision for reconsideration.
- Court
- Administrative Review Tribunal of Australia
- Date
- 2026-08-26
- Who used the AI
- Litigant in person
- AI tool
- Not identified in the record
- Conduct
- A self-represented NDIS applicant's submissions contained multiple fictitious tribunal authorities and inaccurate statutory references prepared with generative artificial intelligence. The Tribunal accepted the AI explanation, warned about the risks under its 2026 practice direction, but decided the merits and remitted the support decision for reconsideration.
- Outcome
- Warning
- Penalty
- No monetary penalty recorded
- Regulatory outcome
- No separate regulatory disposition recorded
- Status
- Verified against listed sources
- Last checked
- 2026-09-04
What was fabricated or misused
- A number of supposed previous AAT decisions cited in the applicant's written submissions — no reported case had the claimed name and citation, or the citation matched a different non-NDIS case; the Tribunal did not list them
- National Disability Insurance Agency v WRMF [2020] FCAFC 79 — cited for a proposition the Tribunal said did not necessarily align with the case
- Section 24 of the NDIS Act — cited as satisfied evidence for an existing participant although it concerns access criteria
What the court said
“The written submissions filed on behalf of the Applicant included reference to a number of supposed previous decisions of the AAT, which were said to be authority for various contentions.”
“Some specific statutory references were also inaccurate. For example, a reference was made to evidence satisfying section 24 of the NDIS Act, which addresses the disability requirements for meeting the access criteria of the NDIS and is no application here given MFRD’s status as an existing participant in the NDIS, who has already met the access criteria.”
“The practice direction continues on to note that the presentation of false and inaccurate information to the Tribunal is unacceptable and may hamper the Tribunal achieving its objectives and making the correct or preferable decision in a proceeding.”
“Secondly, the problem with the fictitious authorities and references infects the Applicant’s written submission to a significant extent, requiring me to effectively disregard some sections in their entirety, and to approach the other parts with considerable circumspection.”
Timeline
- Agency approved MFRD's current plan without core support funding.
- Internal review confirmed refusal to include support worker assistance.
- Mr A lodged an AAT review application for MFRD.
- Review hearing held by Microsoft Teams with Mr A appearing for MFRD.
- Tribunal set aside and remitted the support decision with directions.
Why this case matters
MFRD is different from sanction records such as Mata v Avianca because the AI problem came from a parent appearing for a child in an administrative benefits review. The Tribunal accepted that the tool misled a non-lawyer, but still recorded the practical cost: sections of the submission had to be disregarded or treated with caution. The merits outcome also matters. The applicant obtained some support-worker funding while the reimbursement claim failed. That mix shows why Australian tribunal practice directions focus on disclosure and verification rather than treating every AI error as a disciplinary event.
Practice note
For ART and NDIS matters, a representative using generative AI should check each case name, citation, jurisdiction and statutory section before filing. The check must include whether the case is about the NDIS and whether the cited section applies to the participant’s status. Where the ART practice direction applies, the party should also follow its filing responsibilities for AI-assisted documents.
Primary sources for this incident
Related regulation records for this incident
No related regulation record has been linked to this incident yet.
Similar incidents on the record
- Asif v Minister for Immigration and Citizenship [2026] FedCFamC2G 14022026-07-22 · Federal Circuit and Family Court of Australia (Division 2) · Costs orderDismissing a student-visa judicial review, the Federal Circuit and Family Court of Australia fixed costs of AUD 9,097.93 against a self-represented applicant, increased because of AI misuse, after his ChatGPT-written su…
- Shahin v Industrial Relations Secretary on behalf of Multicultural NSW (No 2) [2026] NSWIRComm 112026-07-21 · Industrial Relations Commission of New South Wales (Full Bench) · OtherA Full Bench of the Industrial Relations Commission of New South Wales refused leave to appeal an unfair-dismissal decision, criticising the self-represented appellant's use of generative AI under Practice Note 33 after…
- Eric Posey v. Summer Bushnell, Docket No. 52072 (Idaho 2026)2026-09-22 · Idaho Supreme Court · WarningIn Idaho, the state supreme court affirmed Eric Posey's defamation judgment after finding Summer Bushnell's pro se appellate briefing relied on fabricated or inaccurate quotations and possibly unverified AI output. The…
- Hancox v Sutherland and others [2026] EAT 1392026-09-17 · Employment Appeal Tribunal · WarningIn the UK Employment Appeal Tribunal, an appellant filed a 300-page ChatGPT skeleton argument without checking accuracy or procedural compliance. No reliance was placed on it at the hearing. The tribunal warned that sim…
Cite this record
SafeLegalAI Global Legal AI Incident Tracker, “MFRD and National Disability Insurance Agency (NDIS), [2026] ARTA 1833”, safelegalai.com/tracker/mfrd-v-national-disability-insurance-agency (accessed 2026-10-07). Data: CC BY 4.0.