Remote AI Adversarial Red Team Specialist

Mercor; New York (NY); Remote; USD 90,000 - 140,000; Full time

Job summary

Mercor is building a remote AI safety red team to probe conversational models and surface vulnerabilities in high-sensitivity topics. You will annotate failures, classify risks, and generate reproducible reports for customers to act on.

Before exposure to content, topics will be clearly communicated, and participation is supported with guidelines and wellness resources. You should bring prior red-teaming experience in AI adversarial work or cybersecurity, be curious and adversarial, structured

Qualifications

• Prior red teaming experience in AI adversarial work or cybersecurity.
• Curious and adversarial: push systems to breaking points.
• Structured: use frameworks or benchmarks, not random hacks.
• Able to explain risks clearly to technical and non-technical stakeholders.
• Adaptable: move across projects and customers.

Responsibilities

• Red team conversational AI models and agents: jailbreaks, prompt injections, misuse cases, bias exploitation.
• Generate high-quality human data: annotate failures, classify vulnerabilities, and flag systemic risks.
• Apply structure: follow taxonomies, benchmarks, and playbooks to keep testing consistent.
• Document reproducibly: produce reports, datasets, and attack cases customers can act on.

Skills

Red team experience

Adversarial mindset

Structured thinking

Communication skills

Adaptability

Tools

Adversarial ML

Cybersecurity

Socio-technical risk

Creative probing

Job description

Mercor is building a remote AI safety red team to probe conversational models and surface vulnerabilities in high-sensitivity topics. You will annotate failures, classify risks, and generate reproducible reports for customers to act on.

Before exposure to content, topics will be clearly communicated, and participation is supported with guidelines and wellness resources. You should bring prior red-teaming experience in AI adversarial work or cybersecurity, be curious and adversarial, structured