Healthcare Eligible Data Breach Assessments: Use case and AI Confidentiality

A results email goes to the wrong person and the practice manager has to decide whether it is an eligible data breach under the Privacy Act. A medical practice that holds patient health information is covered by the notifiable data breaches scheme whatever its turnover. The manager writes an assessment that names the patient, describes the result that was disclosed, counts the 30 days and weighs the harm, then pastes it into ChatGPT or Claude to sharpen the reasoning. That puts a second disclosure of the same patient’s information on top of the first. This page marks every item in a routine assessment record and shows how the same record is drafted on Ermos inside the practice.

The eligible data breach assessment

Opened by the practice manager as soon as the practice is aware of reasonable grounds to suspect an eligible data breach, kept up to date during the assessment and closed with a decision on whether to notify.

Kestrel Lane Medical Practice
Suspected eligible data breach: assessment record · Ref KLMP-PRIV-2026-003 · Opened Wednesday 4 November 2026 · Updated Monday 9 November 2026

What happened.

What was disclosed.

How it came to light.

Awareness and deadline.

Steps taken.

Likelihood of serious harm.

Interim view (9 November).

Telling the patient.

If the decision is to notify.

Fixes.

An invented assessment. The patient, the recipient and the IT contractor are fictional. Hover or tap any highlighted item for the analysis; the full list is below.

Breach assessment record organiser

An Agent+ that helps keep an eligible data breach assessment in order. It builds a timeline from the incident notes, emails and the practice's own response plan, lists what is known and unknown, and flags the questions still open. It drafts updates to the assessment record as facts come in.

What it does

  1. Reads the incident notes, emails and response plan
  2. Builds a dated timeline of what is known
  3. Lists open questions and facts still unconfirmed

What stays with your people. The practice manager, with the principal, decides whether the breach is eligible and whether to notify. The agent organises the record; it does not decide.

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 practice

  1. 1
    The event and two patients

    Names the affected patient with his date of birth, the person who received the email and what caused the mistake. Information about a person’s health is sensitive information, and the recipient’s name is personal information about someone who did nothing wrong. A practice that provides a health service and holds health information is not a small business operator (Privacy Act s 6D(4)(b)), so the Act covers it whatever its turnover. Privacy Act guide.

  2. 2
    The result that was disclosed

    A laboratory result with a hepatitis B finding against a named patient. It is health information under s 6FA and sensitive information under s 6(1). Repeating it in the assessment record, and then in an outside tool, discloses it again. Patient data guide.

  3. 3
    The recipient’s account

    What the recipient said she did and her written confirmation. It is the evidence the practice relies on, written by a patient of the practice who is not the person affected.

  4. 4
    The awareness date and the 30 days

    Section 26WH(2) of the Privacy Act requires an entity that is aware of reasonable grounds to suspect an eligible data breach to carry out a reasonable and expeditious assessment and to take all reasonable steps to complete it within 30 days after becoming aware. The rule is public. The date ties it to this event. Privacy Act guide.

  5. 5
    Steps taken and a named contractor

    Records the containment steps and the IT contractor that carried them out. APP 11.1 requires an entity that holds personal information to take reasonable steps to protect it from unauthorised access or disclosure, and APP 11.3 says those steps include technical and organisational measures. Privacy Act guide.

  6. 6
    The likelihood of serious harm

    Section 26WE(2) describes an eligible data breach as unauthorised access to or disclosure of information that a reasonable person would conclude would be likely to result in serious harm to any of the individuals. Section 26WG lists what to weigh: the kind and sensitivity of the information, whether security measures protected it, who has it or could obtain it, the nature of the harm and any other relevant matters. The weighing is a judgement the practice makes and records. Privacy Act guide.

  7. 7
    The interim view on remedial action

    Section 26WF(1) says that if the entity takes action before the disclosure results in serious harm and, as a result, a reasonable person would conclude serious harm is not likely, the event is not an eligible data breach. Whether the facts meet that test is a decision for the practice, with legal advice where unsure. A draft conclusion is not the decision. Privacy Act guide.

  8. 8
    Telling the patient

    Records a phone call to the affected patient, his reaction and a request for a written account. The patient’s feelings and the chance he will complain sit in the practice’s own record.

  9. 9
    The statement and notification steps

    Section 26WK(3) sets out what a statement must contain: the entity’s identity and contact details, a description of the breach, the kinds of information concerned and recommendations about the steps individuals should take. Section 26WK(2) requires the entity to give a copy to the Commissioner as soon as practicable, and s 26WL requires it to notify affected individuals as soon as practicable after preparing the statement. The steps are public, and sending them to an outside tool exposes nothing beyond the patient’s surname.

  10. 10
    The fixes and the author

    Corrective steps with dates, and the manager who wrote the record. They show what the practice will change and who is accountable for it.

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 practice 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 patient 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 Use cases in healthcare practices. See how this works in practice: Practice policies and forms on Ermos.

Common questions

Do ChatGPT and Claude use the information in a data breach assessment?
Whatever is pasted in 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. This record names Mr Dunstan with his date of birth and his hepatitis B and HbA1c results, names the recipient Ms Dray, and sets out the practice's own view of whether serious harm is likely.
Is it safe to paste a data breach assessment into ChatGPT or Claude?
No, because the assessment repeats the disclosure it is about. 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. Pasting the hepatitis B finding and Mr Dunstan's name into an outside tool discloses them a second time. A practice that provides a health service and holds health information is not a small business operator under section 6D(4)(b) of the Privacy Act, so the Act covers it whatever its turnover.
Does deleting the chat remove a data breach assessment from ChatGPT or Claude?
Not completely. 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). That matters here because Ms Dray deleted the misdirected email and confirmed it in writing, and deleting a chat gives the practice no equivalent confirmation about the assessment.
How can a medical practice use AI for a data breach assessment 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 assessment is drafted from the practice's own incident log and assessment template. Helen Tran and Dr Voss check every date, weigh the harm and make the decision on whether it is an eligible data breach.
Does a GP practice have to notify data breaches?
A practice that provides a health service and holds health information is not a small business operator under section 6D(4)(b) of the Privacy Act, so it is covered. An eligible data breach is unauthorised access to or disclosure of information that a reasonable person would conclude is likely to result in serious harm to an individual. Where the practice has reasonable grounds to believe there has been one, it must prepare a statement for the Commissioner and notify affected individuals (sections 26WK and 26WL).
How long does a practice have to assess a suspected data breach?
Under section 26WH(2) of the Privacy Act, an entity that is aware of reasonable grounds to suspect an eligible data breach may have occurred must carry out a reasonable and expeditious assessment of whether there are reasonable grounds to believe it amounts to one, and take all reasonable steps to complete the assessment within 30 days after becoming aware.
Is a breach involving My Health Record handled differently?
Yes. Section 26WD of the Privacy Act says the notifiable data breaches scheme does not apply to access, disclosure or loss that has been, or is required to be, notified under section 75 of the My Health Records Act 2012. Section 75 applies to registered healthcare provider organisations, among others, and requires notification of the System Operator and the Information Commissioner as soon as practicable after becoming aware.

See Dominion answering from your own documents.

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

Book a demo