AI Safety Experts — English & Portuguese (global)

$29–45/hrrate as listed

Mercor is recruiting bilingual red teamers to probe conversational AI models in English and Portuguese (global varieties, excluding Brazilian Portuguese), documenting jailbreaks, prompt injections, and multi-turn manipulation attempts. Observed pay runs $29–45/hr, remote and asynchronous, with sensitive-content projects clearly flagged in advance and optional.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

You spend your sessions attacking models rather than assisting them. A typical task queue mixes structured probing — working through a taxonomy of harm categories, running scripted attack patterns, and checking whether a model holds its guardrails across a long multi-turn conversation — with open-ended adversarial exploration where you invent a novel jailbreak, role-play framing, or prompt injection and see how far it goes. Language matters here: many safety filters are trained and tuned primarily on English, and attacks that route through European, African, or Asian varieties of Portuguese, code-switching, or translation framing often succeed where the English equivalent fails. Finding and documenting those gaps is a large part of the value you add.

Every finding needs to survive someone else reproducing it. You write up the attack case with the exact prompts, the model's response, the harm classification under the project taxonomy, and a plain-language explanation of why it matters. Reports that read as "the model said something bad" without a reproducible path and a severity rationale tend to get rejected. Work touches bias, misinformation, harassment, and other harmful-behavior categories; topics are communicated before exposure, higher-sensitivity queues are opt-in, and wellness resources are provided.

What the screening looks for

Mercor's process is AI-led: a résumé and profile parse, then a recorded conversational interview, then paid or unpaid sample tasks. The interview probes whether your red teaming experience is real and specific — expect follow-ups asking you to name a framework you've used, walk through an attack you built, or explain why a particular jailbreak worked mechanically. Generic security vocabulary without a concrete case behind it reads poorly. Bilingual fluency is verified conversationally and through written samples; the exclusion of Brazilian Portuguese is a sourcing constraint on this project, not a judgment of the variety, so be explicit about which variety you speak natively.

Logistics

Fully remote and asynchronous; you claim tasks from a queue rather than working fixed shifts.
Contributors commonly report 10–30 hours per week, with volume fluctuating by customer project.
Observed rate is $29–45/hr, positioned by experience depth and language coverage — not guaranteed, and subject to project.
Expect assignment across multiple customers over time; adaptability to new taxonomies and playbooks is part of the job.

What they look for

Native or near-native fluency in both English and Portuguese (global varieties, excluding Brazilian Portuguese)
Hands-on red teaming experience — AI adversarial testing, penetration testing, or socio-technical probing
Working knowledge of LLM attack surfaces: jailbreaks, prompt injection, multi-turn manipulation, bias exploitation
Ability to write reproducible attack reports with clear severity and harm classification
Comfort applying a defined taxonomy or benchmark consistently across many cases
Willingness to review sensitive content covering bias, misinformation, and harmful behaviors

Before you apply

Would you pass this screen?

Drawn from what this listing actually requires. Answered in your browser — nothing is sent to us, and we do not store your answers.

Are you a native speaker of English?
Native fluency in English · required
Are you a native speaker of Portuguese, and is your variant non-Brazilian Portuguese (European, Angolan, Mozambican, or other)?
Native fluency in Portuguese (non-Brazilian) · required
Do you have hands-on experience in at least one of the following: AI red teaming, adversarial testing, cybersecurity penetration testing, abuse analysis, or socio-technical vulnerability probing?
Prior red teaming or adversarial testing experience · required

0 of 3 answered

Included with this listing

Interview prep — 11 questions

Matched to what this platform's screening actually measures. Answer each out loud in 60–90 seconds: claim → concrete example → implication.

What it tests: Whether you have hands-on red teaming or adversarial testing experience, and whether you can move from finding a flaw to producing actionable, reproducible artifacts that non-security teams can understand and fix.
A strong answer: The answer names a specific system (AI model, application, network, or process), describes a concrete attack or probe method you applied (not just 'tested it'), and details how you communicated the finding—ideally with a report, dataset, or reproduction steps that let the defending team fix it without needing you to re-explain.
Where people lose it: Vague stories about 'finding bugs' or 'security testing' without naming the system, technique, or how the finding was documented; or describing finding a flaw but stopping there without explaining how you made it actionable for the team that had to fix it.