AI Red Teamer

Core responsibilities:

• Red-team an unreleased AI product from a frontier lab to identify and mitigate security vulnerabilities
• Apply advanced knowledge of LLM internals and prompt injection techniques to test system boundaries
• Develop jailbreaking strategies for frontier models to expose potential failure modes
• Work independently to drive security testing projects to completion with minimal supervision

This role creates positive change by identifying and mitigating security vulnerabilities in frontier AI systems before deployment, helping prevent potential harms from advanced AI technologies.

Required skills and qualifications:

• Advanced knowledge of LLM internals and prompt injection techniques
• Experience developing jailbreaking strategies for frontier AI models
• Ability to work independently with minimal supervision
• Strong documentation skills for vulnerability reporting
• Collaboration skills to work with AI safety researchers

No benefits information provided in the job description.

Trajectory Labs appears to be an organization focused on AI safety and security, working with frontier AI labs to test and secure advanced AI systems.