Not a giggling matter - when the human in the loop is part of the problem

Giggle for Girls Pty Ltd v Tickle [2026] FCAFC 64

With a name like Giggle and Tickle, the case was always going to attract attention. But even without the giggles this decision is instructive on automated decision-making.  And importantly a cautionary tale for anyone who treats ’human in the loop’ as a compliance talisman.

Giggle for Girls was a women-only networking app. Users were required upload a selfie as part of the registration of their profile.  The third-party AI software then assessed the profile to confirm if the face was that of a woman.

Roxanne Tickle, a transgender woman (whose registration at Births, Deaths and Marriages is female), was admitted to the app by the AI in 2021. Months later, following a human review of her photograph, Giggle revoked her access, and re-admission was refused. Ms Tickle subsequently commenced proceedings with the Australian Human Rights Commission. 

At first instance, Bromwich J found Giggle had engaged in indirect discrimination against Ms Tickle on the ground of gender identity and awarded her $10,000 (Tickle v Giggle for Girls Pty Ltd (No 2) [2024] FCA 960). Both sides appealed. On 15 May 2026 the Full Court (Perry, Abraham and Kennett JJ) dismissed Giggle's appeal, allowed Ms Tickle's cross-appeal, and found Giggle had engaged in two instances of direct discrimination against Ms Tickle, and doubled damages to $20,000 plus limited costs.

The Court held that the exclusion and later refusal to reinstate by Giggle were because of Ms Tickle's gender-related appearance, and that this amounted to direct discrimination under s 22 of the Sex Discrimination Act 1984 (Cth) on the ground of gender identity. The case obviously raises important issues on gender identity; however the focus of this article is the further implications raised by this specific deployment of AI.

The concept of having a ’human in the loop’ has become the reflexive answer to a lot of AI governance. Regulators invoke it, vendors market it, and boards take comfort in it. The assumption is that a human checkpoint launders the automated decision.  Whatever the AI did, a person's judgement cleanses it.

Giggle v Tickle says otherwise, and this shouldn’t be a surprise as the human in the loop is not a disinfectant. Indeed, a human reviewer can introduce the very discrimination the system was meant to prevent. 

For organisations deploying AI, whether in recruitment, underwriting, claims triage, access control or content moderation, the takeaways are uncomfortable but clear. Mapping where in a workflow a decision is actually made is a compliance exercise, not merely an engineering one.

The case name may raise a smile. The issues it raises should sharpen the mind. ’Human in the loop’ is an element of the design, not a defence.  The human added to the loop needs to have clear guidance on the relevant policies, applicable laws, and how to interpret and apply them. A human in the loop is a necessary but not sufficient element of using AI.