Would-be authors were fooled by AI staff and virtual offices in suspected scam (www.theguardian.com)

A network of fake publishing sites operating in Australia, the UK and New Zealand has been using cloned branding, virtual office addresses and AI‑generated “staff” photos and testimonials to dupe aspiring authors into paying upfront fees for vanity-style publishing packages. The Guardian uncovered sites trading as Melbourne Book Publisher, First Page Press, Aussie Book Publisher, Oz Book Publishers and BookPublishers.co.nz; some even reused another publisher’s ABN, listed books actually published by other self‑publishers, and repurposed real authors’ images and names in fake endorsements. Victims reported video calls with convincing interlocutors, small initial payments followed by pricey package offers, and threats when disputes were raised. Australia’s National Anti‑Scam Centre is investigating and affected authors and platforms (Trustpilot, Amazon, Scamwatch) are involved. For the AI/ML community this is a cautionary example of how generative tools amplify social engineering: synthetic faces, paraphrased copy and edited imagery can fabricate credible corporate facades at scale while evading simple manual checks (domain registration records and mismatched metadata were among the red flags). The case highlights urgent needs for technical and policy responses—robust provenance and content‑authenticity signals (watermarks or signed metadata), improved detection models for synthetic media, platform-level verification of business identities, and automated cross-checks that surface mismatches between claimed and registered assets—to reduce adversarial use of generative models.