AI Safety Expert - Adversarial ML

mercor; San Francisco (CA); Hybrid; AUD 22,000 - 30,000; Full time

Job summary

Mercor is seeking AI Safety Experts for English and Odia to remotely test and harden conversational AI models. You will simulate jailbreaks, annotate model failures, and classify vulnerabilities to help clients reduce risks and improve reliability.

This contract role offers flexible hours and the ability to work asynchronously with a global team on guidelines, taxonomies, and playbooks designed for consistent testing and actionable outcomes.

Qualifications

• Native fluency in English and Odia.
• Strong judgment about language and content.
• Rigorous attention to detail and consistency.
• Structured approach to guidelines and quality standards.
• Clear communication with technical and non-technical audiences.
• Adaptability across projects, task types, and customers.

Responsibilities

• Red team conversational AI models and agents. Conduct jailbreaks, prompt injections, misuse cases, and bias exploitation.
• Generate high-quality human data. Annotate failures, classify vulnerabilities, and flag systemic risks.
• Apply structure. Follow taxonomies, benchmarks, and playbooks to ensure consistent testing.
• Document reproducibly. Produce reports, datasets, and attack cases that customers can act on.
• Work independently and asynchronously. Thrive in flexible hours while improving AI model performance.

Skills

English Fluency

Odia Fluency

Judgment

Attention to detail

Structured approach

Communication skills

Adaptability

Adversarial ML

Cybersecurity

Tools

Adversarial ML

Cybersecurity

Job description

About the job

Mercor connects elite creative and technical talent with leading AI research labs. Headquartered in San Francisco, our investors include Benchmark , General Catalyst , Peter Thiel , Adam D'Angelo , Larry Summers , and Jack Dorsey .

Position

AI Safety Experts — English & Odia

Type

Contract

Compensation

$16–$22/hour

Location

Remote

Role Responsibilities

• Red team conversational AI models and agents. Conduct jailbreaks, prompt injections, misuse cases, and bias exploitation.
• Generate high-quality human data. Annotate failures, classify vulnerabilities, and flag systemic risks.
• Apply structure. Follow taxonomies, benchmarks, and playbooks to ensure consistent testing.
• Document reproducibly. Produce reports, datasets, and attack cases that customers can act on.
• Work independently and asynchronously . Thrive in flexible hours while improving AI model performance .

Qualifications – Must-Have

• Native fluency in English and Odia .
• Strong judgment about language and content.
• Rigorous attention to detail and consistency.
• Structured approach to guidelines and quality standards.
• Clear communication with technical and non-technical audiences.
• Adaptability across projects, task types, and customers.

Qualifications – Preferred

• Experience in Adversarial ML : jailbreak datasets, prompt injection, RLHF/DPO attacks, model extraction.
• Background in Cybersecurity : penetration testing, exploit development, reverse engineering.
• Knowledge of socio-technical risk: harassment/disinfo probing, abuse analysis, conversational AI testing.
• Creative probing skills: psychology, acting, writing for unconventional adversarial thinking.