Discovered documents and AI: what the Harman obligation means for ChatGPT or Claude.

Documents obtained from the other side by discovery, subpoena or interrogatories come with a condition most firms learned once and rarely think about: they may be used only for the proceeding in which they were produced. The High Court restated it in Hearne v Street, and it is an obligation owed to the court, with contempt as the sanction. That makes summarising discovered material in ChatGPT or Claude a different problem from ordinary AI confidentiality. This page marks every item in a typical discovery review note and shows how the same review happens on Ermos inside the firm.

Documents obtained on discovery

Prepared by the solicitor reviewing the other side’s discovery, for counsel and the partner, before the documents are used in evidence.

Discovery review note
· Respondent’s List of Documents, categories 2 and 4 · · 20 October 2026

Documents reviewed:

Key documents.

Assessment.

An invented review note. Document numbers are illustrative. Hover or tap any highlighted item for the analysis; the full list is below.

What leaves the firm

  1. 1
    The proceedings

    Named parties, so every document below is tied to a live matter. Rule 9 confidentiality guide.

  2. 2
    The other side’s documents, by number and author

    Three named executives’ internal emails and board papers, produced under compulsion. Under Hearne v Street, a party who obtains documents by compulsory process cannot, without the leave of the court, use them for any purpose other than the proceeding. Legal privilege guide.

  3. 3
    An internal email of the opponent

    A quotation from a document the opponent was compelled to produce. Pasting it into an external tool is a use of the document outside the proceeding, which is the use the obligation to the court prohibits. Legal privilege guide.

  4. 4
    A restructure paper and its date

    The opponent’s own document, undermining its own case. The most valuable discovered document in the matter, and the one a court would least expect to find on a third party’s servers. Legal privilege guide.

  5. 5
    Board papers and an inference about the redactions

    Compelled corporate documents, plus the solicitor’s reading of them. The documents are restricted by the obligation; the reading is privileged. Legal privilege guide.

  6. 6
    A remark about the client’s health by a named executive

    A compelled document, about the client’s leave, by a third party. Restricted, sensitive, and about an identified individual. Privacy Act guide.

  7. 7
    The solicitor’s assessment and the warning to the client

    Privileged advice on the strength of the case, and the very warning about collateral use that the firm has just failed to follow itself. Legal privilege guide.

  8. 8
    The solicitor

    Ties the review to a practitioner personally subject to the obligation, and to the firm. Supervising AI 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 AI data leakage in law firms. See how this works in practice: Discovery on Ermos.

Common questions

Do ChatGPT and Claude use the information in documents obtained on discovery?
Yes, every word pasted in is sent to the AI company's servers to be processed. 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. A discovery note quotes the opponent's emails and restructure paper, produced under compulsion for one proceeding.
Can we use ChatGPT or Claude for documents obtained on discovery if we remove the names?
Not safely, because the obligation in Hearne v Street ties documents obtained by compulsory process to the proceeding, and removing names does not change their origin. 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. The document numbers, verbatim quotations and meeting dates lead straight back to one proceeding. 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 to use for documents obtained on discovery?
Not on its own. The obligation in Hearne v Street limits use of the opponent's documents to the proceeding and is owed to the court. A business plan's terms are agreed with the AI company, so they leave open whether those documents can go there. 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. The OAIC's guidance also tells organisations to review carefully the terms and settings that will apply to their use of any AI product. Our comparisons with ChatGPT Enterprise and Claude weigh the trade-offs.
How can a law firm use AI for documents obtained on discovery without sending them 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 solicitor asks for a chronology in plain language, and it is built from the indexed discovery and the firm's review note template. Each entry cites its source document, and the solicitor checks every citation before signing off the note.
What is the Harman obligation?
The obligation, restated by the High Court in Hearne v Street [2008] HCA 36, that a party who obtains documents or information by compulsory court process cannot, without the leave of the court, use them for any purpose other than that for which they were given, unless they are received into evidence. It is owed to the court and breach can be contempt. It attaches to discovery, subpoenaed documents and interrogatory answers, and extends to third parties who receive the material knowing where it came from.
Is summarising discovered documents in ChatGPT or Claude a breach?
The obligation prohibits use for a purpose other than the proceeding. Whether running the documents through a frontier model to summarise them for the proceeding is such a use has not been decided, and no firm should want to be the test case. What is certain is that the documents have been transmitted to a third party, retained on its terms, and on consumer tiers can be used to train models, none of which the producing party or the court contemplated.
How does discovery review work on Ermos?
The discovered material is indexed on the unit in your office and reviewed there: chronologies, issue summaries and document-by-document notes, each citing the documents it drew on. Nothing is used anywhere except inside the firm, for the proceeding.

See Dominion answering from your own documents.

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

Book a demo