Law Firm Non-Disclosure Agreements: Use case and AI Confidentiality

A non-disclosure agreement is usually the first document in a deal and one of the first a law firm drafts with AI help, because the structure is familiar and the clauses repeat. Staff paste the draft, the client’s instructions and the list of what will be shared into ChatGPT or Claude to tighten the wording. This page marks every item in a realistic mutual NDA between a NSW fabricator and a possible machining subcontractor that rule 9, privilege and the NDA itself bear on, including a clause that restricts generative AI tools, and shows how the same draft is prepared on Ermos.

The non-disclosure agreement

Drafted when a client is about to share commercial or technical information with a possible partner, supplier or buyer, and signed before anything is disclosed.

Mutual non-disclosure agreement
Draft 2 · Carrick Doyle Lawyers for Pacific Rim · Monday 16 November 2026

Parties. Each is a Discloser of the information it discloses and a Recipient of the information it receives.

  1. Purpose.
  2. Confidential Information. All information disclosed for the Purpose, in any form, including .
  3. Obligations. The Recipient must use Confidential Information only for the Purpose, keep it secure with at least the care it gives its own confidential information, and give access only to those who need it for the Purpose.
  4. Generative AI.
  5. Permitted disclosure.
  6. Term.
  7. Return and remedies.
  8. Governing law.
  9. Signed

[Internal, delete before sending]

An invented draft. The companies, the addresses, the certificate numbers and the specification numbers are invented placeholders. Hover or tap any highlighted item for the analysis; the full list is below.

Non-disclosure agreement drafter

An Agent+ that drafts a non-disclosure agreement from the client's instructions about who is sharing what, and with whom. It works from your firm's template, fills in the parties and purpose, and checks the draft for stray deal names, staff names or notes copied from other agreements.

What it does

  1. Reads the client's instructions and the parties' details
  2. Drafts the agreement from your NDA template
  3. Checks for stray names and notes from other matters

What stays with your people. A solicitor reads the draft against the instructions and a partner settles it. The parties sign; the agent drafts only.

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 firm

  1. 1
    The parties and their addresses

    Two companies, and the fact that this client is about to do business with that one. Company details are on public registers. Rule 9.1 of the Australian Solicitors’ Conduct Rules bars a solicitor from disclosing information confidential to a client and acquired during the engagement, and what the client plans, with whom, and that it has asked a law firm to paper it, is part of that until the client chooses to act. Rule 9 confidentiality guide.

  2. 2
    The purpose and the opportunity behind it

    A statement of what the client hopes to win and why it needs a subcontractor. The opportunity has not been announced, and the client has given a promise to another company, in the NDA of 2 September, about it (see item 10). A summary sent to an AI tool discloses the opportunity on the client’s behalf. Rule 9 confidentiality guide.

  3. 3
    The schedule of confidential information

    Named technical documents, a rate card and a competency matrix that names two employees with their certificate numbers. The specifications and rates are the client’s commercial information. The names and certificate numbers are personal information, being about identifiable individuals. Whether the Privacy Act 1988 binds a particular business depends on facts such as its turnover and what it does, and the firm’s duty under rule 9.1 applies either way. Privacy Act guide.

  4. 4
    The generative AI clause

    A contract term that would stop the other side putting the client’s information into a chatbot unless the tool is on its own systems in Australia and does not train a model. Nothing in rule 9 requires such a clause; it is a commercial choice, and the same clause would bind the client for the other side’s information. The wording is the firm’s precedent and holds no client data. The point for this page is the mirror image: a draft NDA that restricts AI tools is itself a document the firm is tempted to paste into one.

  5. 5
    Permitted recipients and the required-by-law carve-out

    Who may see the information: directors, employees and advisers who need it, and anyone the law compels. Rule 9.1 is built the same way for the firm, which may disclose client confidential information to its own partners, principals, directors and employees and to barristers and others engaged by the practice or an associated entity to deliver or administer legal services for the client, otherwise only as rule 9.2 permits, for example with the client’s authority or where the law permits or compels disclosure. The Law Society’s January 2026 guide treats placing client confidential information into a publicly available generative AI system as likely to be a breach of confidentiality and says clients may lose privilege. Rule 9 confidentiality guide.

  6. 6
    Term and survival

    Two years, then three more, and trade secrets for as long as they stay secret. The periods are the client’s commercial choice, and the draft shows how long the client is prepared to be bound.

  7. 7
    Return and remedies

    A promise that copies held “in any tool or system” will be returned or destroyed on request. For information entered into a publicly available chatbot, a user cannot make that promise, because retention and deletion happen on the provider’s terms.

  8. 8
    Governing law

    Standard wording that names a State and its courts. It contains nothing about the client or the deal.

  9. 9
    The signatories

    Two named individuals and their titles. Personal information, and a signature block that the signed NDA would show to both sides anyway; the point is that people with authority sign. Privacy Act guide.

  10. 10
    The internal note

    A restriction in another NDA, the client’s capacity problem, and the unit cost and walk-away price from a different matter. Privileged advice and strategy: s 118 of the Evidence Act 1995 (NSW) prevents evidence being adduced, on the client’s objection, if it would disclose a confidential communication or a confidential document made for the dominant purpose of a lawyer providing legal advice, and whether a particular note qualifies is a question on the facts. Pasting the note discloses a third party’s deal terms and the client’s margin to a company that should not have them. Legal privilege 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 firm 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 matter 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 law firms. See how this works in practice: Contract drafting on Ermos.

Common questions

Do ChatGPT and Claude use the information in a non-disclosure agreement?
Everything 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. This draft carries a competency matrix naming Kofi Mensah, weld inspector, and Ivan Kowalczuk, welder, with their certificate numbers, plus a rail supply contract for 2027 that nobody has announced and an internal note about a different NDA with Meridian Rail Systems.
Can we use ChatGPT or Claude for a non-disclosure agreement if we remove the names?
Not safely, because the Purpose clause still describes one deal. Finish machining of steel frames, a rail rolling-stock supply contract in 2027 and a possible subcontractor point to one client without a company name. 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. Rule 9.1 of the Australian Solicitors' Conduct Rules bars disclosing information confidential to a client, and the Law Society of NSW's January 2026 guide treats placing it into a publicly available generative AI system as likely to be a breach of confidentiality. Section 118 of the Evidence Act 1995 (NSW) can protect the internal note on the Meridian NDA, on the client's objection, if it was made for the dominant purpose of legal advice. 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.
Does a ChatGPT Business or Claude Team plan make it safe for a non-disclosure agreement?
Not by itself. On a personal plan the user cannot choose where conversations are stored or processed. Some business plans let an organisation choose to store data at rest in Australia (OpenAI), which still leaves the content with the AI company. A business plan changes the default on training, but it does not change who receives the content. The OAIC's guidance also tells organisations to review carefully the terms and settings that will apply to their use of any AI product. Clause 4 of this draft allows a generative AI tool only if it runs on systems under the Recipient's sole control in Australia and is not used to train any model, and the client is a Recipient as well as a Discloser. Its staff would need to check any tool against that clause, and against the promise in clause 7 to return or destroy copies held in any tool or system. Our comparisons with ChatGPT Enterprise and Claude weigh the trade-offs.
How can a law firm use AI for a non-disclosure agreement 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 agreement is drafted from the client's instructions and the firm's own NDA precedent. A solicitor checks the list of permitted recipients and the term, confirms that nothing about the Meridian NDA reaches the other side, and a partner approves it before it goes to Jessamine.
Does a non-disclosure agreement stop the other side putting confidential information into an AI tool?
It depends on the wording. A typical NDA lets the recipient pass information only to listed people, such as directors, employees and advisers who need it for the purpose. An AI provider is not on that kind of list, so entering the information into a chatbot is likely to be a disclosure outside it. Some NDAs now say so expressly. Read the clause and the tool's terms before relying on either.
Is a draft NDA confidential to the client the law firm acts for?
Generally yes. Rule 9.1 of the Australian Solicitors' Conduct Rules says a solicitor must not disclose information confidential to a client and acquired during the engagement, except as rule 9.2 permits, for example where the client authorises it or the law permits or compels it. The Law Society of NSW's January 2026 guide says placing client confidential information into a publicly available generative AI system is likely to be a breach of confidentiality.
If a business sends personal information to an overseas AI provider, who is responsible?
For a business bound by the Privacy Act 1988, APP 8.1 deals with disclosure of personal information to an overseas recipient, and under section 16C an overseas recipient's act that would breach the Australian Privacy Principles can be treated as the business's own act, subject to the conditions in that section. Whether the Act binds a particular business depends on facts such as its turnover and what it does.

See Dominion answering from your own documents.

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

Book a demo