There is a specific kind of dread in receiving enthusiastic mail about your work and having to ask: is any living person actually moved by what I made? For writers, that question used to arrive rarely, as an existential shiver. Now it arrives daily, in the form of multi-paragraph emails from Bella Monroe in Barcelona and Thalina Rose in Frankfurt, each one baroque with specific praise — 'deadpan genius,' 'moral courage' — and each one a vector for a wire transfer request. The mechanism that makes these scams newly effective is not simply that AI can write convincing prose. It is that the AI has actually read the books. Anthropic paid $1.5 billion to settle claims from authors whose copyrighted work trained Claude. Google faces a lawsuit for using copyrighted material to train Gemini. The training corpus that enables personalised, plausible fan letters is built on the stolen intellectual labour of the very writers being targeted. The scam is, in a precise sense, the writer turned against themselves. The author received an email purportedly from Ottessa Moshfegh — Booker-shortlisted, one of the more distinctive voices in contemporary American fiction — who described her work-in-progress as 'preoccupied with interior life: how emotions register, how they fail to, and what that absence or excess does to identity over time.' The description is convincing enough that it reads like something Moshfegh might actually say. That's the point. The scam doesn't impersonate badly; it impersonates well, and that is what makes it a genuine literary problem, not merely a spam problem. Rather than simply blocking and moving on, the author played along with a fake 'Berlin Book Club founder,' insisting payment could only be sent in physical cash. After two days of negotiation, a mailing address in the Philippines was produced. A Google Maps search revealed a rural two-lane road with palm trees, a junk shop, an evangelical church, and a large compound hidden behind high white walls — consistent with what investigators have documented as scam farm infrastructure. The Philippines is among the countries most devastated by these operations. A 2024 raid on a compound in Bamban, north of Manila, freed nearly 700 people who had been trafficked and forced to work as scammers — roughly half Filipino, the rest trafficked from China, Vietnam, Malaysia, Taiwan, Indonesia, and Rwanda. A nearby compound contained a brothel, a karaoke suite, and a torture room. Workers sit in sound-insulated booths with a dozen phones each, while supervisors carry tasers. The operations are run largely by Chinese crime syndicates and generate billions annually. The technology these compounds deploy has escalated beyond text generation. Face-masking apps allow video calls conducted behind any chosen identity. The primary con is 'pig butchering' — long-duration romance fraud in which victims are cultivated over months or years before being directed into fake cryptocurrency investments. When Thai forces seized a major Cambodian compound, they found sets for a fake Brazilian police station and a replica Vietnamese bank. A former manager described the work in terms a novelist would recognize: 'It's all about exploiting emotion. You create a dream.' The author reported the address to the Philippines' cybercrime agency. The spokesman for the presidential anti-organised crime commission was candid about the obstacle: the syndicates have 'the financial capability to corrupt just about anybody in public office.' The Bamban compound, when raided, turned out to be run by the local mayor. The expectation that AI companies will voluntarily constrain criminal misuse of their systems faces the same structural problem — it is enormously profitable not to. What started as a story about annoying spam ends somewhere much darker: a global infrastructure of coercion, trafficking, and stolen intellectual property, in which the literary world's most intimate communications — someone saying your work moved them — have become the attack surface.
