Indonesia warns of deepfake scams as AI hoax videos proliferate

Liputan6’s fact‑check is a grounded snapshot of how generative AI is already re‑wiring fraud at street level. The scams described are not futuristic: deepfake videos of ministers announcing fake grants, complete with doctored captions and call‑to‑action phone numbers, are being blasted across Facebook and messaging apps to harvest personal data and lure victims into money‑transfer schemes. The AI here is mostly in the synthetic video and voice, but the consequence is a widening trust gap between citizens and their institutions.

For AGI watchers, this is a reminder that capability progress doesn’t only show up in benchmark scores or developer tools—sometimes it appears first as a wave of cheap, localised fraud. As models get better at language, likeness and style transfer, the cost of producing “good enough” deepfakes targeting specific countries and demographics falls towards zero. That in turn will drive demand for provenance infrastructure, watermarking, and detection models, but also for policy responses like default scepticism of video evidence. In aggregate, this kind of erosion of information integrity could slow down the positive deployment of advanced AI by turbo‑charging backlash and regulatory clampdowns.