NDIS Provider Internal Audit Reports: Use case and AI Confidentiality

Before an approved quality auditor arrives, a service manager reads the participant files, checks the incident and complaint registers and writes down where the service will be marked down. The result is an internal audit report, and it is the kind of long, dull document that gets pasted into ChatGPT or Claude to be tightened, rated against the Practice Standards and turned into an action plan for the board. This one is written twenty-two days before a mid-term audit. It names participants, a fall, a restrictive practice, a worker under supervision, a participant who has opted out of the audit and a staff member’s own paste into ChatGPT. This page marks every item and shows how the same report is drafted on Ermos inside the service.

The internal audit report

Written by the service manager in the weeks before an audit. A registered provider that completed a certification audit and is registered for higher risk or more complex supports completes a mid-term audit, assessed against the standards on provider governance and operational management and any standard that previously needed a corrective action plan. An internal review is the service’s own preparation for the audit. The Quality Management indicators expect a documented program of internal audits, proportionate to the size and scale of the provider and the scope and complexity of its supports.

Brushbox Community Supports · Internal readiness review ahead of the mid-term audit
Tuesday 17 November 2026 · Prepared by Meena Ramanathan, Service Manager · Read by Beth Hargreaves, Team Leader

The audit
Rating key
Risk management
Restrictive practices
Complaints
Human resource management
Participants and the audit
Information management
Actions
Sent to

An invented internal audit report. The case reference and phone number are placeholders. Hover or tap any highlighted item for the analysis; the full list is below.

Internal audit report drafter

An Agent+ that drafts the internal audit report from the audit notes, incident and medication records, and the checklist or standards text you load. It sorts findings into your report layout, shows the record behind each one, and lists gaps that may need a corrective action for the service manager to consider.

What it does

  1. Reads audit notes, records and your checklist
  2. Sets each finding against the text you supply
  3. Drafts the report with the record behind each finding

What stays with your people. The service manager reads, corrects and owns the report and decides what action follows. The agent drafts and cross-checks; it makes no findings of its own.

An Agent+ is built once and anyone in the practice can run it. It works from your own documents and runs on Ermos Dominion, the AI unit in your office, or on Ermos Edge, your own instance in an Australian data centre.

What leaves the service

  1. 1
    The auditor, the audit date and the service’s audit history

    The approved quality auditor’s name and mobile number, the audit start date and the fact that the last certification audit found a minor non-conformity. A registered provider that completed a certification audit and is registered for higher risk or more complex supports completes a mid-term audit, which assesses the governance and operational management standards, any standard that previously needed a corrective action plan and any extra standard the Commission requires. Personal information about the auditor and commercial information about the service. Privacy Act guide.

  2. 2
    The rating key

    The Commission publishes this scale for audit ratings: 3 conforms with elements of best practice, 2 conforms, 1 minor non-conformity, 0 major non-conformity. It names no one and says nothing about this service.

  3. 3
    A fall, a missed step and the service’s own rating

    A named participant, a fall in the shower and the admission that advice from September was not acted on. Health information, which is sensitive information under the Privacy Act. The Risk Management standard says risks to participants, workers and the provider are identified and managed, and the Incident Management standard has incidents acknowledged, responded to, well managed and learned from. A record of a reportable incident is kept for 7 years from the day notification is given (section 25 of the Incident Management and Reportable Incidents Rules). The rating of 1 is the service’s own judgement and not an audit finding. Patient data guide.

  4. 4
    A restrictive practice and a worker’s departure from the plan

    A participant’s PRN medication, a named worker, the minute he phoned the team leader and the service’s view that the use was not in accordance with the plan. Chemical restraint is a regulated restrictive practice (section 6 of the Restrictive Practices and Behaviour Support Rules), and under section 16(3) of the Incident Rules a use that follows a State authorisation is still a reportable incident if it is not in accordance with the behaviour support plan. Health information about Graham and an admission about Tomasz. Patient data guide.

  5. 5
    Complaint numbers, complainants and what they said

    Two open complaints with the names of a sister and a participant, one of them about how a worker spoke about his health. The Complaints Rules require reasonable steps to ensure information provided in a complaint is kept confidential and only disclosed if required by law or if the disclosure is otherwise appropriate (section 8(4)(b)). They also require records of complaints, kept for 7 years from the day the record is made, and statistical information so the provider can review the issues raised (section 10). Privacy Act guide.

  6. 6
    A named worker, an upheld complaint and a late supervision record

    A casual worker’s start date, a complaint upheld against him and the admission that his first supervision record came eight months after he started. The Human Resource Management standard expects workers to be competent in relation to their role. The Code of Conduct covers persons employed or otherwise engaged by an NDIS provider (section 5(2)). Section 7B(3) of the Privacy Act exempts an employer’s act or practice that is directly related to a current or former employment relationship and an employee record it holds about the individual; whether sending his record to an AI company falls inside that is a question for advice. Privacy Act guide.

  7. 7
    A participant who has opted out of the audit

    Sam Ibrahim’s decision not to take part, who he told and what the service will do. The Commission says audit sampling is opt out: the provider tells participants they are automatically included and, if one does not want to take part, respects that decision, documents it and tells the approved quality auditor. The Code of Conduct requires respect for individual rights to self-determination and decision-making and for the privacy of people with disability (section 6(1)(a) and (b)). It sits next to his open complaint in the same report. Privacy Act guide.

  8. 8
    A staff paste into ChatGPT and the lack of a rule

    The service’s own record that a worker put a participant’s details into ChatGPT, and that no written rule exists. For an APP entity, APP 6.1 limits disclosure for another purpose and APP 8.1 applies before it discloses personal information to an overseas recipient. If an APP entity is aware of reasonable grounds to suspect an eligible data breach, it must carry out a reasonable and expeditious assessment and take all reasonable steps to complete it within 30 days (section 26WH(2)); whether this is one is for the service to assess. The Commission’s February 2026 position statement is about behaviour support plans, and says disclosing sensitive participant information to a third-party AI platform such as ChatGPT without safeguards may breach the Code. Privacy Act guide.

  9. 9
    Actions, owners and dates

    Three people, their tasks and deadlines, a trainer’s name and a session time. Low sensitivity alone, but it shows where the service is thin. The Quality Management standard expects a quality management system that promotes continuous improvement of support delivery.

  10. 10
    Who gets the report, and the expected result

    The board chair’s name, the board date and the service’s forecast of its own ratings before the auditors have seen anything. The Governance and Operational Management standard expects each participant’s support to be overseen by governance and operational management systems proportionate to the size and scale of the provider and the scope and complexity of the supports delivered.

What happens in Frontier Models (ChatGPT, Claude)

  • The whole document is transmitted to the provider’s servers. On personal plans the user cannot choose where it is stored or processed, and for the major providers that has generally meant outside Australia, under that country’s legal process. Some business plans offer storage in Australia, but the provider still receives the content.
  • On consumer tiers, prompts can be retained and used to train future models.
  • The provider’s staff and systems can reach the content under the provider’s own policies.
  • Once sent, the service no longer controls it. Retention and deletion happen on the provider’s terms.

The same document on Ermos

Ermos Dominion

The document is drafted on a unit in your office, from your own templates and the participant file. Every marked item stays on the unit, which keeps working with the internet switched off. Nothing in it is transmitted to an AI provider, so there is no disclosure to analyse: the confidentiality work is the same as for your file server.

Ermos Edge

Your documents, the index and your access boundaries stay on your own instance in an Australian data centre. When a staff member asks for this draft, only their question and the excerpts they are permitted to see go to a named Australian provider, who is contractually barred from keeping the content or training any model on it. Nothing is processed outside Australia.

Frontier Models (ChatGPT, Claude)Ermos EdgeErmos Dominion
Where the document is processedProvider servers, generally offshore on personal plansYour instance in an Australian data centre; answers generated in AustraliaThe unit in your office
Who can read itProvider staff and systems, under policyYour staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothingYour staff only
Is it keptOn the provider’s termsKept on your instance; the provider is barred from retaining itStays on the unit
Can it train a modelOn consumer tiers, yesNo, by contractOnly your own unit learns from it, on your hardware

Part of Use cases in NDIS services.

Common questions

Do ChatGPT and Claude use the information in an internal audit report?
Yes. Everything pasted into ChatGPT or Claude is sent to the AI company's servers and processed there to produce a reply. As at October 2026, whether a conversation on a personal plan (ChatGPT Free, Plus or Pro; Claude Free, Pro or Max) can be used to train the models depends on a setting each user controls. Business plans (ChatGPT Business and Enterprise, Claude Team and Enterprise) and the developer APIs sit under commercial terms, and OpenAI and Anthropic each say they do not train on that content by default (OpenAI, Anthropic). On either product, the thumbs up and thumbs down buttons can send a whole conversation back for training. A readiness review pasted in to be tightened carries the lead auditor's name and mobile number, Graham's fall in the shower and the admission that the September advice was not on his risk assessment, and Sam Ibrahim's decision not to be interviewed.
Can we use ChatGPT or Claude for an internal audit report if we remove the names?
Not safely, because the report still describes real people without their names. The Office of the Australian Information Commissioner (OAIC) recommends in its guidance on commercially available AI products that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools, because of the significant and complex privacy risks. Removing names rarely makes a document anonymous. Information stays personal information while the person is reasonably identifiable, and the OAIC warns that information can be at risk of re-identification even when it has been de-identified or anonymised. A man who fell in the shower on 20 October, a casual worker supervised eight months after starting and a complainant whose complaint concerns a worker talking about his health at a clinic front desk are easy to place for anyone who knows the service.
Can ChatGPT or Claude be trusted to draft an internal audit report?
Not without a check. Language models can produce fluent text with details that are not in the source, wrong figures or dates, and references that look real but are not. A manager should check every item against the registers and files. The most harmful errors here would be the complaint count and average closing time, the dates of the fall and the PRN tablet, and the ratings, because the board and the auditor may rely on them.
How can an NDIS provider use AI for an internal audit report without sending it to ChatGPT or Claude?
Ermos is built so the document stays under your control. On the Dominion AI unit it is drafted inside your office and none of it is sent to an AI provider. On Ermos Edge your documents and index stay on your own instance in an Australian data centre, and only the question and the excerpts the user is permitted to see go to a named Australian provider that is contractually barred from keeping them or training any model on them. The draft is produced from the participant files, the incident and complaint registers and the service's own review template. The service manager checks every count, date and rating and signs the report before it goes to the board.
What is an NDIS mid-term audit?
The NDIS Commission says registered providers that completed a certification audit and are registered to provide higher risk or more complex supports must do a mid-term audit, which has to start no later than 18 months after the registration period begins unless the Commissioner allows longer. It assesses the standards on provider governance and operational management, any standard previously needing a corrective action plan and any extra standards the Commission requires. The provider engages an approved quality auditor, and some provider types are excluded.
Do participants have to take part in an NDIS audit?
No. The Commission describes audit sampling as opt out. A provider must tell participants they are automatically included, which can mean an interview or a review of their file, records or plan. If a participant does not want to take part, the provider must respect that decision, document it and tell the approved quality auditor.
How long must an NDIS provider keep records of reportable incidents?
Section 25 of the Incident Management and Reportable Incidents Rules requires a registered NDIS provider that becomes aware of a reportable incident to keep a record of it for 7 years from the day notification of the incident is given. The Rules note that other Commonwealth, State or Territory laws may also apply to the retention of records.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo