What the work involves

You write prompts in Italian designed to probe how a model behaves on sensitive subject matter — self-harm, extremism, weapons and dual-use chemistry, medical and legal advice, fraud, harassment, election and civic misinformation. Then you turn around and apply a structured taxonomy to prompts and multi-turn conversations, marking where a response crosses a policy line and where it stays inside one. A large part of the job is written justification: the label alone is worth little to the client, and reviewers read your reasoning as closely as your classification.

The Italian dimension is the point. Adversarial phrasing that reads as obviously hostile in English often arrives in Italian as dialect, euphemism, regional slang, deliberate misspelling, or a formal register that lends a harmful request an air of legitimacy. You are expected to catch escalation patterns that build across turns rather than appearing in a single message, and to know when something is genuinely regionally specific — Italian legal terminology, local scam conventions, political framing — versus a straight translation of an English-language pattern.

What the platform screens for

Mercor's screen is an AI-led interview, generally 20–40 minutes, sometimes with a short Italian writing sample. Expect probing on: whether your Italian is truly native or near-native under fast, idiomatic follow-up; whether your written English is clear enough to carry policy reasoning to an English-reading reviewer; whether you can hold a consistent line across borderline cases instead of drifting; and whether you understand where the line sits between information that is uncomfortable and information that is genuinely operational. Candidates who treat every sensitive topic as automatically disallowed screen out about as often as candidates who wave everything through.

Logistics

  • Fully remote and asynchronous; no fixed shifts, though task batches can carry deadlines.
  • Immediate start; contributors commonly commit 15–30 hours per week, with part-time arrangements typical.
  • Observed pay band is $40–44/hr — reported by contributors, not a guarantee.
  • Based in Italy or Western Europe is a stated preference only; applicants elsewhere are explicitly welcome.
  • A bachelor's degree, completed or in progress, is required. Direct AI/ML experience is not.
  • Sustained exposure to distressing subject matter is part of the role; consider that honestly before applying.