The crypto asset market
attracts growing segments of retail investors who lack the tools to
assess risk. Fraud losses in 2024 exceeded $5.8 billion, with a marked
increase in scams facilitated by generative AI. This work proposes an
operational taxonomy of frauds: honeypots (contractually embedded
on-chain), rug pulls (post-fundraising), and pig butchering
(relationship-based). Particular attention is devoted to cases in which
tokens generated massive losses without meeting clear criminal law
thresholds. For each category, documentary, on-chain, and behavioural
signals are identified to support preventive risk assessment,
acknowledging that defences remain structurally backwards-looking
relative to the innovation of fraudulent actors.
