YISROEL MIRSKY

A messaging scam is one in which a scammer exploits a victim through a dialogue of text messages, a process that can unfold over hours, weeks, or even months. Scams such as romance-baiting ("pig butchering"), family-emergency impersonation, account-security impersonation, and sextortion all depend on sustained text-based dialogue to build trust or fear and coach victims through irreversible actions. Through an insider investigation of crime syndicates in Southeast Asia, we have observed that LLMs are increasing both the scale and quality of these scams, a dangerous trend that has received remarkably little attention. To help bring awareness to this threat and expose its emergence, we (1) discuss how LLMs are industrializing these crimes, (2) provide a taxonomy of messaging scams amenable to LLM operation, and (3) discuss findings from our recent insider investigation of crime syndicates in Southeast Asia, in which 145 insider interviews (2022-2026) revealed routine, institutionalized use of ChatGPT across scam compound operations. These interviews expose organizational structures that concentrate labor in early-stage messaging, with ChatGPT serving as a standardized tool for tone polishing, translation, and context-aware reply drafting. We conclude that mitigation cannot rely on per-message detection alone and motivate defenses centered on long-horizon behavioral signals, public awareness, and out-of-band verification norms.