Reportable incident reports and AI data security: what an NDIS incident notification gives away in ChatGPT or Claude.

A serious injury or an alleged neglect has to be notified to the NDIS Commission within 24 hours of the service becoming aware of it. That is exactly when a service manager, already shaken, reaches for ChatGPT or Claude to help write a clear and complete account. The draft names a participant with a brain injury, a worker who stepped out of the room, a neighbour who heard the fall and the service’s own failings. This page marks every item in a draft notification that the Incident Management Rules, the Code of Conduct and the Privacy Act bear on, and shows how the same report is drafted on Ermos inside the service.

The reportable incident notification

Prepared by the service manager when a reportable incident occurs in connection with supports. For a serious injury, abuse or neglect, the notification goes to the Commissioner within 24 hours of the service becoming aware (section 20 of the Incident Management and Reportable Incidents Rules), with witnesses and further actions within 5 business days. The service keeps a record of the incident for 7 years from the day notification is given (section 25).

Reportable incident notification (draft for the NDIS Commission portal)
Brushbox Community Supports · Prepared by · Wednesday 21 October 2026, 7.40am

Person with disability.

Incident type.

Basis.

What happened.

The worker.

Witness.

Immediate actions.

Contributing factors.

An invented notification. The NDIS number is a placeholder. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the service

  1. 1
    The participant

    A named man with a brain injury, his address and his NDIS number. Health information and personal information that the Information Management standard expects to be kept confidential. Patient data guide.

  2. 2
    The category and the clock

    That a serious injury and alleged neglect occurred at a named service, and exactly when the service became aware. The 24 hours run from the service becoming aware, so this line also records whether the service was on time.

  3. 3
    The notification rule

    Section 20 of the Incident Management and Reportable Incidents Rules, in general terms. The same for every registered provider and published by the regulator.

  4. 4
    The account of the fall

    A step-by-step account of an injury to a named man, including the worker’s movements. Health information, and an allegation that has not yet become a finding. Patient data guide.

  5. 5
    The worker and a personal circumstance

    A named worker, a sick child at school and his stand-down. His personal information and an allegation about his conduct, which the Code of Conduct and the service’s own review process exist to deal with fairly. Privacy Act guide.

  6. 6
    A witness

    A neighbour named in a report to the regulator. Her personal information, in a document she did not write. Privacy Act guide.

  7. 7
    Actions taken and who was told

    Section 20 requires the notification to say what immediate actions were taken, including whether police or any other body has been told. It also names the nominee, the support coordinator and a team leader. Privacy Act guide.

  8. 8
    What the service had not yet changed

    An admission that a recommended safeguard was not yet in place and a worker was not yet signed off. The service’s own account of contributing failures. The Commissioner can require a final report with investigation details, and the service keeps the record of the incident for 7 years.

  9. 9
    The manager

    Ties the report to the responsible person. Members of the key personnel must take all reasonable steps to ensure reportable incidents are notified (section 18 of the Rules).

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 AI data leakage in NDIS services.

Common questions

Do ChatGPT and Claude use the information in a reportable incident notification?
Yes. A draft pasted into either tool 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. The draft names Graham Pettit and his fractured wrist, a worker who left the bathroom to take a phone call and a neighbour who heard the fall.
Is it safe to paste a draft incident notification into ChatGPT or Claude?
No, because the draft joins health information and an allegation in one document. 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. Without Graham's name it still describes a 58-year-old man with an acquired brain injury, a fall in the shower on a named date and a casual worker who started in September. 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.
Does deleting the chat remove an incident notification from ChatGPT or Claude?
Not completely, so a deleted draft should not be treated as gone. Deleting a chat removes it from the user's history, but OpenAI and Anthropic each say they can keep a deleted chat on their systems for up to 30 days, and longer where safety or legal obligations apply (OpenAI, Anthropic). The service itself keeps its record of a reportable incident for 7 years from the day notification is given, under section 25 of the Incident Management and Reportable Incidents Rules, so a draft that left the service is a second copy it does not control.
How can an NDIS provider use AI for a reportable incident notification 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 incident entries on the participant file and the service's own notification template. The service manager checks every time, name and action against the file and submits it within the 24 hours the Rules allow.
What is a reportable incident for an NDIS provider?
Under the NDIS Act, a reportable incident is the death of a person with disability, serious injury, abuse or neglect, unlawful sexual or physical contact or assault, sexual misconduct including grooming, or the use of a restrictive practice other than in accordance with a State or Territory authorisation. Alleged incidents count. The Rules narrow some of these, for example by excluding negligible unlawful physical contact.
How quickly must a reportable incident be notified?
A registered provider must notify the Commissioner within 24 hours of becoming aware of a death, serious injury, abuse or neglect, unlawful contact or assault, or sexual misconduct. Other reportable incidents, such as an unauthorised restrictive practice, must be notified in writing within 5 business days. The Commissioner may later require a final report within 60 business days of the notification.
How long must the service keep the incident record?
Seven years from the day notification is given, under section 25 of the Incident Management and Reportable Incidents Rules. The Rules add that other Commonwealth, State or Territory laws may also apply to record retention. A record kept for seven years is a reason to decide early where a draft report may go, because a copy on a third party's servers is a copy outside the service's control.

See Dominion answering from your own documents.

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

Book a demo
Keep reading
AI data leakage in NDIS services Resources The Privacy Act, the APPs and AI All professions