AI Safety Expert

About the role

The AI Safety Expert will join Mercor’s team in Dubai to help secure conversational AI models and agents. You will focus on identifying jailbreaks, prompt injections, misuse cases, and bias exploitation while generating high‑quality human data for testing.

Key responsibilities

• Red‑team conversational AI models, targeting jailbreaks, prompt injections, and bias exploitation.
• Generate, annotate, and classify failure data, flagging systemic risks and documenting reproducible test cases.
• Produce reports, datasets, and attack scenarios that customers can act on.
• Review AI outputs on sensitive topics such as bias, misinformation, or harmful behavior.

Required profile

• Native fluency in English and Malayalam with strong judgment about language and content.
• Rigorous attention to detail, ability to spot subtle errors, inconsistencies, and gaps.
• Excellent communication skills to explain reasoning to both technical and non‑technical audiences.
• Adaptability to work across diverse projects, task types, and customers.

Required skills

• Adversarial ML knowledge: jailbreak datasets, prompt injection, RLHF/DPO attacks, model extraction.
• Cybersecurity experience: penetration testing, exploit development, reverse engineering.
• Understanding of socio‑technical risk: harassment/disinformation probing, abuse analysis, conversational AI testing.
• Creative probing abilities: psychology, acting, unconventional adversarial thinking.