Businesses that are parties to contractual agreements (and what business is not?), are beginning to consider the risk and potential consequences of the other party inputting their business’s confidential information into public AI systems.

In Australia, cases where the cause of action is breach of confidence arising in this particular circumstance may not yet be apparent in the courts. However, this may only be a matter of time, and when they surface, the courts will be ready to deal with these issues. Australian law already comprehensively addresses confidentiality and the protection of sensitive information in a variety of contexts. And already Courts both here and overseas have directly addressed misuse of AI (including through breach of confidence) by both lawyers[i] and self-represented litigants[ii], and via guidance issued.[iii]

Court treatment of confidentiality loss via AI use

One context in which Australian law and the courts have addressed the protection of sensitive information and AI use is family law. Relevantly in the 2025 case of Mertz v Mertz (No 3), the Court deplored the actions of three lawyers who had used public AI systems to research and prepare submissions.  The Court explained that:

…in proceedings under the Family Law Act 1975 (C’th) (the Act), by s 114Q of the Act and subject to various exceptions, communication to the public of an account of proceedings that identifies a party, a witness or an associated or concerned person is prohibited. There is a risk that entering draft documents into an AI program will result in a breach of s 114Q, a breach of the Harman undertaking, breaches of rules in respect of material produced under subpoena and/or give rise to a waiver of legal professional privilege.[iv]

Overseas, in 2026 in the matter of United States v Heppner, a man charged with securities fraud had retained lawyers but also separately used Claude-AI to develop strategy and submission documents, which he provided to the lawyers. The man’s assertion of client legal privilege (legal professional privilege or LPP) over those documents failed for three reasons, including that confidentiality had not been maintained due to the man’s disclosures to AI.[v]

Contractual protection

In the broader context of a commercial contract then, where one or both parties may have little understanding that entering either confidential information disclosed to them or the terms of the agreement itself into a public AI system can jeopardise confidentiality, standard confidentiality clauses may not offer adequate protection against risk.

A specific clause is recommended to make clear the parties’ separate or mutual obligations in respect of AI and machine learning, including whether:

  • the processing of confidential information using AI is prohibited altogether, or
  • the processing of confidential information using AI is permitted under strict conditions, with standard exceptions.

If the information to be kept secret is highly sensitive or critical to the business, additional clauses may be required, for example to indemnify the other party for any losses arising out of a breach of the confidentiality obligations or to notify if breach is suspected.

Although it can be difficult to enforce confidentiality undertakings, the inclusion in agreements of specific and reasonable AI clauses relating to defined scope of information to be protected is likely to be a key factor in determining enforceability.

Expert legal advice can assist you to reduce your confidentiality risk via tailored AI clauses.

[i] Mertz v Mertz (No 3) [2025] FedCFamC1A222 at [14] ’…relevant duties of legal practitioners, [include] the duty not to mislead the Court, to deliver legal services competently and diligently and not to engage in conduct which is likely to diminish public confidence in the administration of justice or bring the legal profession into disrepute.’

[ii] Re inappropriate use of AI by self-represented litigants: in 2025 the NSW Court of Appeal explored this issue in depth in May v Costaras,[ii] a property dispute:  ‘The respondent’s written and oral submissions were a cogent demonstration that the use of artificial intelligence by non-legally trained users is likely to add to the cost and complexity of legal proceedings without appreciable benefit.’ Per Payne JA at [49]. See also Bell CJ at [2]-[17] in relation to AI use by litigants: May v Costaras [2025] NSWSC 90. See also JML Rose Pty Ltd v Jorgensen (No 3) [2025] FCA 976 (19 August 2025) re hallucinated case references, legislation and court rules.

[iii] In NSW for example, see Practice Note SC Gen 23 – Use of Generative Artificial Intelligence (Gen AI).

[iv] Mertz ibid at [15].

[v] United States v Heppner, No. 25 Cr. 503 (JSR), (SDNY), Memorandum of Rakoff J, 17/02/2026.

For further information please contact:

This article is for general information purposes only and does not constitute legal or professional advice.  It should not be used as a substitute for legal advice relating to your particular circumstances.  Please also note that the law may have changed since the date of this article.