Incident reports and AI security: what a SIRS notification gives away in ChatGPT or Claude.

A serious incident report is written under time pressure. Priority 1 incidents must be notified to the Aged Care Quality and Safety Commission within 24 hours, and the report has to be clear, complete and accurate. That is exactly when a manager reaches for ChatGPT or Claude to help write it, and exactly the document that must not go there: it names a resident, an alleged perpetrator, witnesses and the provider’s own failings. This page marks every item in a reportable incident notification that the Aged Care Act 2024 and the Privacy Act protect, and shows how the same report is drafted on Ermos inside the service.

The serious incident report

Prepared by the facility manager when a reportable incident under section 16 of the Act occurs or is alleged, for notification to the Commission within the required timeframe.

Reportable incident notification (draft for the Commission portal)
Kurrajong Gardens Aged Care · Prepared by · 23 October 2026, 11.10am

Incident type.

Person affected.

What is alleged.

How it came to light.

Immediate actions.

Contributing factors.

Police.

An invented report. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the service

  1. 1
    The resident

    A named man with dementia, his diagnoses and his room. Health information and personal information the Act requires the provider to protect. Patient data guide.

  2. 2
    The incident category

    That a Priority 1 reportable incident has occurred at a named service. Protected information under section 21 of the Act, and reputationally sensitive before the Commission has seen it.

  3. 3
    The allegation and the worker

    A named worker accused of assaulting a resident. His personal information, an allegation that is not yet a finding, and the agency’s name. If this paragraph surfaced anywhere, it would be defamatory of him or devastating for the resident, depending on the truth. Privacy Act guide.

  4. 4
    The witness

    A named worker who reported a colleague. Whistleblowers are protected under the Act; their identity on a third party’s servers is the opposite of protection. Privacy Act guide.

  5. 5
    Actions taken and the son’s earlier complaint

    The provider’s response, and a family member’s statement that he had raised concerns before. Information that would matter in any investigation or claim. Privacy Act guide.

  6. 6
    What the service got wrong

    Missed induction, short staffing and a behaviour support plan not followed. The provider’s own admission of contributing failures, written for the regulator.

  7. 7
    A police event number

    That police have been notified, with the reference. Information about a criminal investigation.

  8. 8
    The manager

    Ties the report to the responsible person, who has her own obligations under the Act.

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 resident’s record. 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 aged care providers.

Common questions

Do ChatGPT and Claude use the information in a serious incident report?
Yes, a pasted incident report is sent to the AI company and processed there to produce the 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. It names the resident, the worker alleged to have used force and the staff member who reported it.
Can we use ChatGPT or Claude for a serious incident report if we remove the names?
No, because the incident itself identifies the people, with or without their names. 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 room number, a date and time, bruising seen at 8.15am, an agency's name and a police event number each narrow the report to one resident and one worker. 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.
Can ChatGPT or Claude be trusted to draft a serious incident report?
Not without the facility manager checking every line against the incident record. Language models can write fluent text that contains details missing from the source, wrong figures, wrong dates and references that look genuine and do not exist. The most damaging errors would be a wrong incident category or priority, a wrong time for the alleged force or the bruising, and a wrong police event number. The account of what the service did next must also match the record. A Priority 1 report is due to the Aged Care Quality and Safety Commission within 24 hours, which leaves little time to find an invented detail.
How can an aged care service use AI for a serious incident 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 manager asks for a draft built from the incident entries, the care plan and the roster in the service's own template. Each statement is tied to its source, and the manager checks and signs the report.
What is a reportable incident under the Aged Care Act 2024?
Section 16 lists them: unreasonable use of force; unlawful sexual contact or inappropriate sexual conduct; psychological or emotional abuse; unexpected death; stealing from or financial coercion of the person by a worker; neglect; use of a restrictive practice other than in accordance with the rules; and unexplained absence. The Commission's current guidance requires Priority 1 incidents to be notified within 24 hours and Priority 2 within 30 days.
The report goes to the Commission anyway. Why does the draft matter?
The Commission is the regulator, bound by law in how it handles the report. A frontier model provider is a company with servers, for the major ones outside Australia, retaining the draft on its own terms. The resident, the accused worker and the witness are all identified in a document that was only ever meant for the regulator.
How would Ermos help with an incident report?
The incident entries, the care plan and the roster are on the unit in your own building, and the report is drafted from them, with every statement tied to the record it came from. The 24-hour deadline is met without the report leaving the service.

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 aged care providers AI for healthcare AI and patient data All professions