Task: given one suspicious message, output a validated JSON object with a
risk level, scam type, manipulation tactics, a compact "scam DNA" breakdown, a
safest action, a message to forward to a trusted person, and a short summary.
Languages: English (en) and Russian (ru).
Nature: synthetic / template-generated suspicious messages and
benign controls. These are not collected real scams (see Limitations).
value
Rows
2,324 (1,200 en + 1,124 ru)
