The orphanage that never existed: an influencer uses AI to fabricate charity and solicit donations

By ABC News · July 5, 2026.

The investigation by the ABC NEWS Verify team reconstructs a case that is worth reading carefully because it condenses, in a single file, much of the risk that generative AI brings with it in the short term. Lily Jay, an Australian influencer who converted to Islam and has almost three million followers on Instagram, founded the Lily Jay Foundation and, since September of last year, has published content about supposed humanitarian works: an orphanage in Uganda, a bakery in Gaza, distribution of Qurans, construction of mosques. The problem is that much of that material is fiction generated by AI: the very Lily Jay who appears announcing the opening of the orphanage is a synthetic version, the children with lollipops are made up, and even the foundation's logo on the workers' T-shirts has the typographical errors typical of these systems. The Ugandan government registry found no orphanage registered under that name, and ABC NEWS Verify found no independent trace of its existence.

The case becomes even more revealing with the 'humanitarian award' the foundation says it received in May: the images of the presentation carry the SynthID watermark of OpenAI's tools, technical proof that they were generated by AI. The press release announcing the award came from an agency, Real Media Group, which on its own website presents Lily Jay as 'co-founder' while treating her as a 'client' in the statement; the site disappeared shortly after ABC contacted one of the directors named. The foundation, moreover, admits in the fine print of its site that it is not a charity but a 'private commercial firm' whose contributions are not tax-deductible, something most of its donors are probably unaware of. After receiving the outlet's questions, the site redirected Australian visitors to a version with no option to donate, but kept the international payment channel active, verified by the journalists using a VPN.

As sector context, this type of fraud is not an isolated anomaly but the most visible manifestation of a trend we have already been highlighting: generative AI is industrializing fraud and disinformation with an ease and a cost that were previously unthinkable. No army of forgers is needed, nor a production budget: a convincing video of smiling children in a fictional orphanage is generated in minutes and reaches millions of people thanks to the organic reach of an account with a loyal audience. The exploitation of charity is nothing new—as the former director of World Vision Australia recalls in the article, images of children in orphanages have always tugged at the donor's heartstrings—but generative AI multiplies the scale and reduces the marginal cost of manufacturing that emotion to almost zero.

Our reading is that this episode clearly illustrates the uncomfortable side of the transition we have been describing for months: while technology advances toward scenarios of abundance and disease prevention, in the immediate arena it also multiplies the tools for deception, and the victims are usually those who act in good faith. The encouraging thing, however, is that the very infrastructure that allows fraud to be detected—watermarks like SynthID, journalistic verification units like ABC NEWS Verify, forensic analysis of visual inconsistencies—is maturing in parallel with generative capability. It is no coincidence that the most conclusive evidence against the foundation comes from a watermark embedded by the model maker itself: the race between generation and detection, just as happens in cybersecurity, is beginning to have institutional counterweights.

In the medium term, the episode should accelerate two things the charity sector has needed for some time: mandatory, visible registration verification before accepting online donations, and pressure on platforms like Instagram to require accounts used for fundraising to have a minimum level of traceability. Trust is the most valuable—and most fragile—asset of any legitimate NGO, and each case like Lily Jay's erodes that collective capital for organizations that do operate transparently. The underlying lesson is not to distrust AI itself, but to demand that verification evolve at the same pace as the ability to manufacture reality.