AI Safety Experts – English, Marathi

. Red-team conversational AI models and agents using jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation .

Posted 9/25/2026contractRemote • United StatesMid-LevelSenior💰 $16 - $22 per hourWebsite

Tools & technologies

Cyber Security

Key responsibilities & impact

Red-team conversational AI models and agents using jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation
Generate human data by annotating failures, classifying vulnerabilities, and flagging systemic risks
Apply taxonomies, benchmarks, and playbooks to maintain consistent testing
Produce reproducible reports, datasets, and attack cases for customers
Review AI outputs involving sensitive topics such as bias, misinformation, and harmful behaviors
Uncover vulnerabilities automated tests miss
Expand evaluation coverage and reduce production surprises
Help Mercor customers improve the safety, robustness, and trustworthiness of their AI systems

What you’ll need

Native fluency in English and Marathi is required
Strong judgment about language and content; ability to assess whether AI responses are accurate, complete, and appropriate
Ability to identify subtle errors, inconsistencies, and gaps
Ability to follow guidelines and quality standards consistently
Ability to explain reasoning clearly to technical and non-technical audiences
Adaptability across projects, task types, and customers
Independent contractor status
Must not require H1-B or STEM OPT support
Nice-to-have: adversarial ML experience, including jailbreak datasets, prompt injection, RLHF/DPO attacks, or model extraction
Nice-to-have: cybersecurity experience, including penetration testing, exploit development, or reverse engineering
Nice-to-have: socio-technical risk experience, including harassment/disinformation probing, abuse analysis, or conversational AI testing
Nice-to-have: psychology, acting, or writing experience for unconventional adversarial thinking

Comp & perks