Australia’s generative-AI boom has produced feel-good stories about productivity tools, medical imaging and agricultural sensors. It has also produced something far darker, and a new investigation by the ABC has put a Sydney startup at the centre of it. According to the ABC’s reporting, the company has quietly built a highly profitable business generating AI pornography, and its tools have raised serious questions about the creation of child sexual abuse material.
The story lands at an awkward moment for a local industry keen to be seen as a responsible player in the global AI race. It also tests, in the most confronting way possible, whether Australia’s much-vaunted online safety regime can actually hold a homegrown AI company to account.
The context: a booming, barely policed corner of AI
AI image and video generators have improved at a startling pace over the past two years. The same diffusion models that let a small business owner mock up a logo, or a teacher illustrate a lesson, can be fine-tuned to produce photorealistic nudity on demand. A cottage industry has grown up around that capability: platforms that host custom models, sell credits for image generation, and let users type a prompt and receive explicit imagery in seconds.
Much of that market operates in a legal grey zone. Adult content involving consenting adults is lawful in Australia, but the technology does not neatly distinguish between an adult and a child, or between a willing participant and a real person whose likeness has been scraped without consent. Child safety experts have warned for more than a year that open-source image models can be pushed, deliberately or through weak guardrails, into generating synthetic child sexual abuse material, which is illegal in Australia regardless of whether a real child was involved.
That is the backdrop against which the ABC’s investigation should be read. The publication reports that a Sydney-based operation has scaled this kind of AI pornography into a genuine business, and that the way its systems work invites hard questions about whether abuse material can be produced through them.
The news
Per the ABC, the startup has built what the outlet describes as an “AI porn empire”, a set of products successful enough to generate real revenue, run from Australia rather than an offshore jurisdiction chosen to dodge scrutiny. The most serious element of the reporting is not that adult AI content exists, but the allegation that the underlying technology raises the prospect of child abuse material being generated, and the difficulty of guaranteeing that it cannot be.
FluentSea has not independently verified the ABC’s findings, and this article reports on that investigation rather than reproducing it. The company has not been named here because the identifying details and any response it may have provided sit within the ABC’s own reporting, which readers can assess directly via the link above.
Two ways to see it
To child-safety advocates and many technologists, a locally run business of this kind is exactly the scenario regulators were warned about. Their argument is straightforward: if a model can produce explicit imagery of adults on demand, the same pipeline can be coaxed toward illegal material, and “we added a filter” is not a sufficient defence when the tooling, the hosting and the profit are all Australian. In this view, the location of the business is a feature, not a footnote, it means Australian law clearly applies, and Australian authorities have no excuse not to act.
The other view, put forward by parts of the AI-development community, is more uncomfortable to sit with but worth stating plainly. Builders of generative tools often argue that they cannot fully control what users do with general-purpose technology, that safety classifiers catch the overwhelming majority of prohibited prompts, and that punishing a domestic operator simply pushes the same activity offshore, out of reach of Australian courts and beyond the sight of Australian investigators. Whether that argument amounts to genuine harm-minimisation or convenient deflection is precisely what an investigation like the ABC’s is designed to surface.
Both positions can be true at once: guardrails are imperfect, and imperfect guardrails are not a licence to profit from a system that can generate the worst material imaginable.
The Australian stakes
This is where the story stops being an abstract debate about AI ethics and becomes a test of Australian institutions. The eSafety Commissioner already has some of the strongest online-safety powers in the world, including the ability to demand the removal of illegal content and to hold platforms to enforceable industry codes. The Commissioner’s office has repeatedly flagged AI-generated abuse material as an emerging threat and has pushed the industry toward “safety by design”, building protections in before a product ships, not bolting them on after harm occurs.
A domestic company changes the enforcement calculus. Where offshore platforms can ignore Australian notices, a Sydney business has directors, bank accounts, cloud contracts and a physical footprint within reach of the law. Producing or distributing child sexual abuse material, including wholly synthetic material, is a serious criminal offence under Commonwealth and state law. If the ABC’s questions are borne out, this becomes a matter for police and prosecutors, not just regulators.
There is a reputational dimension too. Australia is trying to position itself as a trusted place to build and deploy AI, courting investment and talent while governments overseas tighten their own rules. A local “AI porn empire” tangled up in child-safety concerns is the kind of story that follows an industry around, and it will sharpen calls for clearer obligations on anyone training, fine-tuning or hosting generative image models on Australian soil.
What’s next
Expect three things to move. First, pressure on the eSafety Commissioner and law-enforcement agencies to say publicly whether they are examining the operation described by the ABC. Second, renewed debate over whether Australia’s forthcoming AI guardrails, the government has signalled mandatory obligations for “high-risk” AI, should explicitly capture synthetic sexual imagery and the tools that enable it. Third, scrutiny of the infrastructure layer: the payment processors, cloud providers and app stores that quietly keep these businesses running, and whether they should be doing more to cut them off.
For the wider Australian AI sector, the uncomfortable lesson is that “we’re just a platform” is wearing thin. As generative tools get more capable, the distance between a legitimate image generator and an abuse-material factory narrows to a few prompts and a missing safeguard. Investigations like this one are how that gap gets exposed, and, ideally, closed.
Sources: ABC News.
