NDIS Provider Participant Risk Assessments: Use case and AI Confidentiality

Before a new support starts, a registered NDIS provider works through the risks with the participant: what could go wrong, how likely it is and what the worker will do. The quality indicators expect risk assessments to be done in collaboration with each participant, documented in the support plan and reviewed. Team leaders paste their notes into ChatGPT or Claude to turn them into a tidy risk matrix. The notes describe a young woman with Down syndrome, her health, how she shows pain, her wish to swim in the deep end, strangers she might follow and the change room. This page marks every item in an NDIS community access risk assessment that the Practice Standards and the Privacy Act bear on, and shows how the same assessment is written on Ermos inside the service.

The participant risk assessment

Completed by the team leader with the participant and her nominee before a new support starts, and reviewed after the first few shifts. The quality indicators for the Support Planning standard expect risk assessments to be undertaken regularly in collaboration with each participant and documented in their support plans, to consider how far the participant relies on the service and what disruption would mean for her health and safety, and the effectiveness of the risk strategies to be reviewed periodically with her.

Brushbox Community Supports · Participant risk assessment
Community participation, Saturdays · Completed Thursday 5 November 2026 at home with Mei Lin and Thu Tran by Beth Hargreaves, Team Leader · Review after four shifts, by Monday 7 December 2026

Participant
Health
Communication
Rating method
Water (medium)
Strangers and crowds (medium)
Change room (low)
Money (low)
If a shift is cancelled
Agreed and shared

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

Risk assessment pre-fill agent

Built once as an Agent+, it pre-fills a participant's risk assessment from the referral, intake notes and any earlier plans, using your own form. It carries over health, communication and activity notes, leaves strategy fields open, and later compares shift notes against the strategies written.

What it does

  1. Reads the referral, intake notes and earlier plans
  2. Pre-fills your risk assessment form from them
  3. Compares later shift notes with the strategies written

What stays with your people. The team leader assesses risk with the participant and her nominee and writes the strategies. The agent pre-fills and flags; it does not assess.

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 participant

    A young woman’s name, age, address, NDIS number and household, and who her nominee is. Personal information the Information Management standard expects to be identifiable, accurately recorded, current and confidential. Privacy Act guide.

  2. 2
    Disability and health

    Down syndrome, a thyroid condition, hearing loss and how long she can swim. Section 6FA of the Privacy Act counts information about an individual’s health, including a disability, as health information, and health information is sensitive information. Patient data guide.

  3. 3
    How she shows she is unwell

    The signs a worker must watch for because she may not say she is in pain. The quality indicators for the Safe Environment standard expect clear arrangements so workers understand a participant’s communication needs and how she expresses emerging health concerns. Health information, written for her safety. Patient data guide.

  4. 4
    The rating method

    A likelihood and consequence matrix that appears in every assessment the service writes. No one is named in it.

  5. 5
    Water safety and her choice

    Her wish to swim in the deep end and the controls agreed with her. The quality indicators for the Access to Supports standard say access to supports will not be withdrawn or denied solely on the basis of a dignity of risk choice the participant has made. A decision about her, made with her, that the service has to be able to explain.

  6. 6
    Strangers and getting separated

    That she may hug or follow people she has just met. A vulnerability the Code of Conduct duty to take all reasonable steps to prevent exploitation and abuse of people with disability exists to protect, and something nobody outside her support needs to read. Privacy Act guide.

  7. 7
    The change room arrangement

    How she dresses, where the worker waits, the worker’s name and a request about her mother. The Privacy and Dignity standard expects supports that respect and protect her dignity and right to privacy. The most private part of her Saturday, written down. Privacy Act guide.

  8. 8
    Money

    How much she carries and that she sometimes gives money away. The Participant Money and Property standard says participant money is secure and each participant uses her own money as she determines. It also tells anyone who reads it that she can be asked for money.

  9. 9
    Reliance on the service

    The Support Planning quality indicators expect risk assessments to consider how far a participant relies on the service for daily living needs and how her health and safety would be affected if supports were disrupted. Here the answer is low, which lets the service plan around a missed shift.

  10. 10
    Agreement and who may see it

    With consent, the quality indicators expect support plans to be communicated to the support network, other providers and relevant government agencies, and the Information Management indicators expect consent to cover the purpose of a disclosure. Her agreement covers workers rostered to her. An AI company is not on that list. Privacy Act guide.

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 a participant risk assessment?
Yes. Anything pasted or uploaded 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. Notes pasted in to build a risk matrix carry Mei Lin Tran's Down syndrome, her thyroid tablet and the signs that she is unwell. They also record her wish to swim in the deep end and the controls agreed with her.
Is it safe to paste a participant risk assessment into ChatGPT or Claude?
No, because the assessment pairs sensitive information about a young woman with the situations where she is most exposed. 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. Section 6FA of the Privacy Act counts information about a person's health, including a disability, as health information, which is sensitive information. Her Down syndrome, thyroid condition and hearing loss all fall within it. The assessment also says she may hug people she has just met or go where they go, and that she sometimes gives money away. Mei Lin and Thu agreed it may be shared with workers rostered to her, and an AI company is not on that list.
Can ChatGPT or Claude remember the details in a participant risk assessment later?
Both products also have memory features that can carry details from one conversation into later ones, and on a personal account each user decides whether they are switched on (ChatGPT, Claude). That matters here because a team leader writes a risk assessment before each new support, with the same likelihood and consequence matrix every time. Mei Lin's change room arrangement, her friendliness with strangers and the $30 she carries for lunch are details her agreement limits to workers rostered to her. A team leader would not want them turning up in a chat about the next participant.
How can an NDIS provider use AI for a participant risk 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 risk matrix is drafted from Beth's notes, the participant file and the service's own risk assessment form. Beth goes through each rating and control with Mei Lin and Thu, and signs the assessment once they agree it.
Do registered NDIS providers have to do participant risk assessments?
For registered providers assessed against the core module of the Practice Standards, which covers supports such as assistance with daily personal activities and community participation, the Support Planning quality indicators, which must be taken into account in those assessments, expect risk assessments to be undertaken regularly in collaboration with each participant and documented in their support plans, with strategies to treat known risks. They also expect the assessment to consider how far the participant relies on the service and how disruption would affect their health and safety, and periodic reviews of whether the strategies work.
What does dignity of risk mean under the NDIS Practice Standards?
The quality indicators for the Access to Supports standard say each participant is supported to understand when supports can be withdrawn, and that access to supports will not be withdrawn or denied solely on the basis of a dignity of risk choice the participant has made. The Person-centred Supports standard also requires supports that respect individual rights to self-determination and decision-making.
Does the participant have to be involved in a risk assessment?
The Support Planning standard says each participant is actively involved in developing their support plans, and its quality indicators expect risk assessments to be done in collaboration with each participant and reviewed with them. Under the quality indicators a nominee can give consent on the participant's behalf, but the participant's own views still shape the controls, such as how close a worker stays in the pool.

See Dominion answering from your own documents.

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

Book a demo