Large Language Law

May v Costaras [2025] NSWSC 90 concerns Michael May (”the appellant”) and Lila Costaras (”the respondent”) who had been in a relationship since late 2020 living together in various properties, including one owned by the appellant at Cumberland Reach and later in Queensland.

In June 2022, they purchased an investment property in Scott Street, Maryborough, QLD, as joint tenants for $180,000, with the full purchase price paid from the appellant’s bank account. The relationship ended in September 2022. The appellant sought a declaration in the NSW Supreme Court that the respondent held her interest in the Scott Street property on trust solely for him. 

Lindsay J instead found a joint endeavour constructive trust (following Muschinski v Dodds and Baumgartner)—an investment project to benefit both parties, considering their past, present, and future contributions. Their “joint endeavour” ended prematurely when their relationship broke down, leaving unresolved claims to beneficial ownership. 

Although the appellant paid the full purchase price and later excluded the respondent from the property, this did not extinguish the respondent’s equitable interest, particularly given the substantial financial and labour contributions (over $50,000) to renovations of the earlier Cumberland Reach property. 

Equity required recognition of the respondents contributions, the appellant’s sole financial outlay, and the fact that the respondent was denied the opportunity to enjoy or profit from the investment. Lindsay J held it would be unconscionable for the plaintiff to deny the defendant any interest, but equally unconscionable for the defendant to claim a full half share. A constructive trust was declared, with the appellant entitled to two-thirds and the respondent to one-third, and orders proposed to sell the property and division of net rental income in the same proportions.

Appeal

In May v Costaras [2025] NSWCA 178, the appellant submitted that he should have received 100% (or at least 95%) of the property. The appeal raised issues including:

• Whether Lindsay J should have found a resulting trust in the appellants’ favour.
• Was a constructive trust based on common intention wrongly applied?
• Whether the joint endeavour extended beyond Scott Street to include the Cumberland Reach property.
• Whether imposing a constructive trust based on a failed joint endeavour was correct.
• Whether the two-thirds/one-third division was an error.
• Whether findings disregarded contemporaneous records and failed to address allegedly false credit card entries by the respondent.

Primary Judge’s Findings on Evidence

Lindsay J preferred the respondent’s version of key conversations, viewing the appellant’s account as a reconstruction shaped by the romance and its breakdown.

The appellant’s belief that the property was bought “on condition of marriage” was seen as narrow and retrospective.

While acknowledging that the respondent may have fabricated some financial records, Lindsay J found the respondent an honest and reliable witness who relied on the appellant’s promises to commit to the relationship and provide for her children.

Court of Appeal Decision

The Court of Appeal (Payne JA delivering the leading judgment, with Bell CJ concurring separately and McHugh JA agreeing) dismissed the appeal.

The appellant’s first ground—that Lindsay J erred by not applying the presumption of resulting trust (based on his sole payment of the Scott Street purchase price) or by not treating the presumption of advancement as rebutted—was rejected.

Key Findings

Resulting trust not applicable: Lindsay J correctly held that presumptions of resulting trust or advancement did not determine the case but were determined by the particular facts. Even if the presumption of resulting trust arose, it was rebutted.

Intention at purchase: Lindsay J found that by placing the Scott Street property in joint names, both parties intended to hold the property beneficially as co-owners from the outset.

“Conditional on marriage/renovation” rejected: The appellant argued that the respondent’s interest in marriage or renovations was conditional. Lindsay J found this lacked certainty and did not reflect the reality of the relationship. There was no concrete marriage plan, and the renovation condition was unsupported.

Conflicting evidence: In the June 2022 conversations, Lindsay J preferred the respondent’s version, which reflected her reliance on promises of recognition for financial and non-financial contributions, particularly to renovations. Lindsay J held that the appellant’s account was a reconstruction influenced by the relationship’s breakdown.

Broader context: The respondent’s evidence better aligned with the facts—she sacrificed her interests, supported the appellant, and sought recognition for her role in their joint property ventures.

No factual error: The challenge to the primary judge’s acceptance of the respondent’s evidence failed; the applicant showed no reviewable error under Fox v Percy or Lee v Lee.

The Court of Appeal held that the parties purchased the Scott Street property as joint tenants in recognition of financial and non-financial contributions, with a shared intention to restore it for mutual benefit. Accordingly, the Court of Appeal dismissed ground 1 of the appeal and found the presumption of resulting trust inapplicable or rebutted.

Constructive Trust – Failed Joint Endeavour

The Court confirmed that Lindsay J found a constructive trust based on a failed joint endeavour, not a common intention trust. Although the judgment occasionally referred to “common intention,” the analysis relied on the principles in Muschinski v Dodds and Baumgartner. These required proof of:

1. A joint endeavour,
2. It’s a breakdown without attributable fault, and
3. What equity demands are between the parties?

The appellant ultimately conceded in oral submissions that this was the correct characterisation. As a result, grounds 2 and 3 fell away.

Grounds 4 and 5 – Scope of the Joint Endeavour

Ground 4: The appellant argued that Lindsay J failed to clarify whether the joint endeavour included only the Scott Street property or the Cumberland Reach and Ann Street properties. The Court of Appeal rejected this. 

It was common ground that Scott Street was part of the failed joint venture. Further, Lindsay J found the respondent made significant financial and non-financial contributions to renovations and household expenses at Cumberland Reach and Ann Street. Her contributions extended well beyond domestic duties—she resigned from employment to manage renovations and undertook skilled work. Thus, her efforts supported the parties’ broader material welfare.

Ground 5: The appellant asserted Lindsay J wrongly confined the joint endeavour to Scott Street. The Court of Appeal rejected this as unsound, noting that the findings recognised the respondent’s contributions across multiple properties.

Outcome

The Court of Appeal held that the constructive trust was properly characterised as a failed joint endeavour involving Scott Street (and informed by contributions at Cumberland Reach and Ann Street). Therefore, grounds 2, 3, 4, and 5 were rejected.

Constructive Trust – Common Purpose and Contributions

Lindsay J imposed a constructive trust on the Scott Street property because:

• The parties acquired it for a shared purpose – a joint investment, to renovate and resell.
• This purpose was tied to their joint benefit, considering past, present, and future contributions.
• Both parties’ contributions were considered: the appellant’s towards acquisition/renovation of the Cumberland Reach property, and the respondent’s substantial financial and non-financial contributions to those renovations.

Ground 6 – Complaints About the Constructive Trust

6(a) – No pleaded joint endeavour:

• The appellant argued that no one alleged the joint endeavour relied upon.
• The Court of Appeal held that the issue of how the case was run was squarely raised.
• No unfairness resulted, and Lindsay J was entitled to impose a constructive trust. Rejected.

6(b) – Failure to consider appellant’s contributions:

• The appellant submitted that his contributions at Cumberland Reach were overlooked.
• The Court of Appeal found Lindsay J expressly considered these when describing the parties’ common purpose. Rejected.

6(c) – No findings on respondent’s payments:

• The appellant argued Lindsay J failed to identify which payments counted as contributions.
• The Court of Appeal noted Lindsay J accepted the respondent had made substantial contributions (over $50,000), even if precise figures could not be determined, supported by credibility findings. Rejected.

6(d) – Overlooked reimbursements:

• The appellant said some contributions had been reimbursed (e.g., $10,300).
• The Court of Appeal found that Lindsay J had made a finding of contributions exceeding $50,000; the appellant did not challenge this on appeal. Rejected.

6(e) – No findings on benefits received:

• The appellant argued Lindsay J did not determine whether contributions benefitted him financially.

The Court of Appeal explained that under Muschinski and Baumgartner, non-financial contributions can justify relief, and “common benefit” need not be financial (Joudo v Joudo [2024] NSWCA 258 cited).

• The judge correctly identified contributions (financial, household, managerial) without needing to prove a direct economic benefit. Rejected.

Outcome

The Court held that Lindsay J properly imposed the constructive trust based on a failed joint endeavour and that the primary judge had sufficiently addressed both parties’ contributions. Ground 6 in all its parts was rejected.

Constructive Trust – Proportions of Ownership

The Court of Appeal noted that Lindsay J’s reasoning drew on the principles in Muschinski v Dodds, Baumgartner v Baumgartner and Joudo v Joudo. The respondent’s significant financial and non-financial contributions to the joint endeavour were central to the outcome.

Although the appellant had paid the entire purchase price of the Scott Street property, the respondent’s proven contributions to renovations at Cumberland Reach, her expectation of recognition, and the parties’ common plan to restore Scott Street justified her inclusion as co-owner. Lindsay J found it unconscionable for the respondent to retain a full half-share, but equally unjust to deny her any beneficial interest.

The Court of Appeal upheld the allocation of two-thirds to the appellant and one-third to the respondent, rejecting the appellant’s argument for either full ownership or a 95% share. The orders reflected a fair balancing of contributions. They were moulded “to the justice of the case” in an orthodox application of equitable principles.

Ground 7 was rejected.

Grounds 8 and 9 – Factual and Credit Findings

Ground 8 – Contemporaneous records

• The appellant argued that Lindsay J erred by making findings at [108]–[110] without proper regard to contemporaneous written records.
• Counsel for the appellant accepted this was effectively a claim that the findings were “glaringly improbable” under Fox v Percy.
• The Court of Appeal rejected this. Lindsay J had the advantage of hearing both parties give evidence and concluded the respondent was the “better historian,” whose account of the critical conversation was more reliable.
• The text messages relied on by the appellant actually supported Lindsay J’s conclusion. In particular, a message of 20 June 2022 showed the respondent referring to her contributions and need for security, consistent with Lindsay J’s findings.
• Ground 8 dismissed.

Ground 9 – Falsified credit card entries

• The appellant argued Lindsay J failed to make findings about whether the respondent herself created false credit card statements.
• The Court of Appeal held the complaint was baseless. Lindsay J had already made a serious credit finding: falsification of about $5,000 worth of entries. As the respondent did not explain, this damaged her credibility—at least regarding the extent of her financial contributions.
• The appellant conceded this was a serious credit finding. The Court noted no material difference between creating false records and knowingly relying on them. The judge was entitled to weigh all the evidence when assessing credit.
• Ground 9 dismissed.

Generative AI and Oral Submissions

Bell CJ (Payne and McHugh JJA agreeing) dismissed the appeal but made separate observations about the respondent’s use of generative AI in preparing her oral submissions.

• The self-represented respondent candidly admitted she had used AI to draft her submissions. While some parts were coherent, many were unintelligible, irrelevant, or based on non-existent cases.
• Examples included references to Lipkin Gorman v Karpnale, In re Nisbet and Potts’ Contract, and Wheatley v Bell, none of which were relevant to the appeal.
• The citation of Tate v Ragg [2004] NSWCA 306 was more troubling. It was a fabricated case with false pinpoint references illustrating the risk of AI hallucinations—apparently credible but fictitious authorities.
• The Chief Justice emphasised he was not criticising the respondent personally. Still, the case highlights the risks of unverified AI use, particularly by unrepresented litigants.

Broader Principles and Comparative Authority

• Bell CJ endorsed the recent warnings in Ayinde v The London Borough of Haringey [2025] EWHC 1383 (Admin), which stressed that while AI can support access to justice, practitioners must use oversight and verification.
• Freely available AI tools cannot conduct reliable legal research. They may produce confident but false assertions, fabricate authorities, or misquote real ones.
• Lawyers have a professional duty to check AI outputs against authoritative sources (official databases, law reports, reputable publishers). Leadership in the profession, as well as regulators, must ensure compliance.
• Misuse of AI has already led to erroneous submissions in several jurisdictions (cases cited included Olsen v Finansiel Stabilitet A/S [2025]; Zzaman v HMRC [2025]; Burragubba v State of Queensland [2016]; Mohareb v Saratoga Marine Pty Ltd [2020]; Vernon v Bosley (No 2) [1999]).
• Even unrepresented litigants are obliged not to mislead the court (Barton v Wright Hassall LLP; Mohareb). Courts may legitimately inquire whether AI has been used to manage cost, complexity, and misinformation risks.

Key Takeaway

This case underscores the need for judicial vigilance in the face of generative AI use in litigation. While AI may improve access to justice, its current risks—particularly fabrication of authorities and irrelevant arguments—require careful oversight. Litigants must verify the accuracy and relevance of all AI-assisted submissions, in line with Practice Note SC Gen 23 (NSW).