AI Deepfake Child Sexual Abuse Material Laws

The advent of AI has introduced revolutionary changes that have advanced society across many fields. However, some of these changes have been in the wrong direction.

It is widely acknowledged that sexual offences are especially destructive for societies at large, so courts in Australia have punished new manifestations of sexual deviancy as they arise, in every form they arise. As such, courts and the broader public are now grappling with sexual crime committed through AI, and the issue is rapidly developing.

Despite the perverse and illegal nature of child abuse material, AI now makes it easier than ever to create – all that may be needed is a blurry picture, a social media username, or even nothing at all. These images are often called ‘deepfakes’.

This sinister paradox has made headlines nationally, after a Queensland school teacher allegedly used AI to conjure explicit images of students and staff. The incident has displayed how our law targets this conduct, and its limitations when faced with offending that evolve so quickly.

The School Teacher

In late July 2026, a teacher at a private school in Arundel, on the Gold Coast, was charged by Queensland Police after he allegedly used AI software on his work-issued laptop to generate sexually explicit content of students and staff. The school, A.B Paterson College, is a co-educational school for students ranging from Prep to Year 12.

He has since had his employment terminated and is set to face court soon.

The school’s principal has stated that it appears the material was not distributed or shared by the accused.

Queensland Child Abuse Material Laws and Penalties

The schoolteacher had been charged with offences pursuant to the Criminal Code Act 1899 (Qld). The relevant provisions are section 228B (involving a child in making child exploitation material or making child exploitation material), section 228D (possessing child exploitation material), and section 408E (misuse of a restricted computer, which essentially encompasses devices protected by a password).

Commonwealth offences of child abuse material under section 474.22 and 474.23 of the Criminal Code Act 1995 (Cth) may also apply where material is accessed, transmitted, or possessed using a carriage service, which captures most AI image-generation tools that run through cloud-based servers. This is a technical area of law that is misunderstood, since many are unaware how wide the scope of a “carriage service” is.

Importantly, Queensland’s definition of child exploitation material does not necessitate that the material depicts or reflects a ‘real event’. This means that generated image of what merely appears to be a child is enough – it does not need to be rooted in reality any further than that.

Commonwealth legislation apply the same logic. They define child abuse material by considering whether the material depicts a person who is, appears to be, or is implied to be a child.

Again, the nature of how the material was created or if it depicts an actual child is irrelevant. So, AI-generated content falls under the scope, since the law is concerned with the existence of child abuse material itself.

Child Abuse Material Using AI Deepfake Laws and Penalties in New South Wales

The alleged offences occurred in Queensland, but had they occurred in New South Wales, a notably different legislative scheme would apply. Here, this sort of offending is codified into two separate regimes: one addressing deepfakes of children, and another addressing adults.

The relevant NSW offence is section 91H of the Crimes Act 1900 (NSW), which criminalises the production, dissemination, or possession of child abuse material, carrying a maximum penalty of 10 years’ imprisonment.

The definition of “produce” in section 91H expressly includes altering or manipulating any image for the purpose of making child abuse material: this precisely captures AI-generated or digitally fabricated content. It is not an element of the offence that the material depicts a real incident of abuse, only that it depicts what a reasonable person would regard as offensive in the way the section describes, involving a person who is, appears to be, or is implied to be a child. The bar here is very low.

This means an entirely synthetic image, generated from a real student’s face but with no underlying “real” abuse ever occurring, is treated with the same gravity as material derived from an actual assault.

Common defences under section 91HA, such as innocently possessing the material or receiving it unsolicited, would have little success on facts such as these, where the material was intentionally created by the accused.

Where the alleged victims are adults (in this case, the school staff), prosecutors will instead be armed with NSW’s newer intimate image offences, which have recently been refurnished to specifically target the rise of adult AI deepfakes.

After the Crimes Amendment (Intimate Image and Audio Material) Bill 2025 (NSW) and the earlier Crimes Amendment (Deepfake Sexual Material) Bill 2024 (NSW), the definition of “intimate image” in section 91N has been broadened to expressly capture images that have been digitally created, altered, or generated to depict a real, identifiable person’s private parts or sexual activity. This applies even where no genuine underlying image ever existed.

New offences, such as the recent section 91PA and the additions to section 91R, now specifically target the production of wholly AI-generated sexually explicit material depicting a real person. These offences are punishable by up to 3 years imprisonment.

These laws work towards closing a dangerous gap in NSW closing what was previously a significant gap, where “revenge porn” laws in 2017 were built around the non-consensual sharing of genuine recordings, but not artificial ones. As misconduct using AI increased, so did the necessity to accurately punish these emerging offences.

The Objections to These Laws

The rapid expansion of legislation that targets deepfake AI-generated material strikes at the core of something obviously dangerous but has still garnered significant criticism.

Civil liberties advocates have raised concerns that some drafting is broad enough to capture parody, satire, or artistic commentary involving public figures, without adequate exceptions. Nudity and lewdness are often used as powerful artistic features, and some wonder whether new legislation may accidentally impugn innocent people expressing their creativity.

Aside from the legislation, many have raised concerns regarding the enforcement of these laws. Offenders frequently use VPNs, overseas-hosted AI tools, or encrypted platforms, which makes detection and prosecution difficult even where the law is clear. As such, the issue may not be in poor legislative drafting, but in poor investigative capabilities.