AI Safety Experts — English & Finnish

$48–68/hr

36 people hired

RemoteGlobalTechnology3-5 YearsContractPosted Aug 11, 2026

About this role

AI Safety Expert responsible for red teaming conversational AI models and agents to identify vulnerabilities, unsafe behaviors, bias, misinformation risks, and adversarial weaknesses. The role involves designing and executing structured adversarial tests including jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation. Experts will generate high-quality human evaluation data, classify vulnerabilities, document reproducible attack cases, and provide actionable findings that improve AI safety and robustness. Native-level fluency in both English and Finnish is required.

AI Red Teaming, Prompt Injection, Jailbreak Testing, AI Safety Evaluation, English & Finnish

Responsibilities

Red team conversational AI models and agents using jailbreaks, prompt injections, misuse scenarios, bias exploitation, and multi-turn manipulation.
Identify, annotate, and classify AI failures, vulnerabilities, and systemic safety risks.
Apply established safety taxonomies, benchmarks, frameworks, and testing playbooks.
Create reproducible attack cases, evaluation datasets, and detailed safety reports.
Evaluate sensitive AI outputs involving bias, misinformation, harmful behavior, and socio-technical risks.
Communicate identified vulnerabilities and recommendations clearly to technical and non-technical stakeholders.

Qualifications

Native fluency in English and Finnish.
Prior experience in AI red teaming, adversarial AI testing, cybersecurity, or socio-technical probing.
Strong understanding of jailbreaks, prompt injection, adversarial testing, and conversational AI vulnerabilities.
Ability to conduct structured testing using frameworks, benchmarks, and taxonomies.
Strong analytical thinking and ability to identify vulnerabilities automated testing may miss.
Excellent written communication and reproducible technical documentation skills.
Experience with adversarial ML, penetration testing, RLHF/DPO attacks, abuse analysis, or AI safety evaluation is highly preferred.