AI Safety Red Team Expert, English and Vietnamese

Help strengthen conversational AI systems by probing them with adversarial inputs, identifying vulnerabilities, and creating high-quality safety data. This text-based role focuses on finding risks that automated testing may miss and turning findings into actionable reports, datasets, and attack cases. Key Responsibilities Red team conversational AI models and agents using jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation. Review AI outputs involving sensitive topics, including bias, misinformation, and harmful behaviors. Annotate model failures, classify vulnerabilities, and flag systemic risks. Use established taxonomies, benchmarks, and playbooks to conduct consistent testing. Create reproducible documentation and artifacts that support improvements to AI systems. Qualifications Native fluency in both English and Vietnamese is required. Prior red teaming experience in AI adversarial work, cybersecurity, or socio-technical probing. Ability to use structured frameworks and benchmarks, communicate risks to technical and non-technical audiences, and adapt across projects and client needs. Preferred Specialties Adversarial machine learning, including jailbreak datasets, prompt injection, RLHF or DPO attacks, or model extraction. Cybersecurity, including penetration testing, exploit development, or reverse engineering. Socio-technical risk assessment, including harassment or disinformation probing, abuse analysis, or conversational AI testing. Creative adversarial thinking informed by psychology, acting, or writing. Work Terms Remote, hourly position. All work is text-based. Participation in higher-sensitivity projects is optional. Topics will be communicated before exposure, with clear guidelines and wellness resources available. Compensation 17 to 25 per hour. Impact Expand evaluation coverage across more scenarios and reduce unexpected production risks. Deliver reproducible findings that help make AI systems more robust, safe, and trustworthy. #J-18808-Ljbffr SaidGig