Conventional Weapons Safeguards Analyst

Socket.dev; New York (NY); Hybrid; USD 245,000 - 330,000; Full time

Job summary

Anthropic is seeking a Safeguards Enforcement Analyst focused on Conventional Weapons to detect and mitigate misuse of our AI systems, applying policy enforcement, threat analysis, and red-team style evals across a range of dangerous technologies.

You will design scalable review workflows, develop evals, and partner with Engineering and Data Science to close gaps, while remaining aware that you may encounter explicit content related to violence or harm.

Qualifications

• Deep, applied expertise in weapons systems and the ability to translate technical evidence into enforcement decisions.
• Experience in policy enforcement, threat intelligence, counterterrorism, or related fields with exposure to harmful content or dangerous technology.
• Experience scaling policy enforcement or content review workflows.
• Proficiency in SQL and/or other data analysis tools to draw insights from large datasets.

Responsibilities

• Design and architect automated enforcement systems and scalable review workflows.
• Develop and maintain evals to measure model performance on policy areas and surface regressions.
• Partner with Engineering and Data Science to optimize detection and enforcement systems for violations.
• Review flagged content to drive enforcement decisions and surface gaps in policy.

Skills

Weapons systems

Policy enforcement

Workflow design

SQL

Threat intelligence

Generative AI

Cross-functional collaboration

Education

Bachelor’s degree or equivalent

Tools

SQL

Job description

Anthropic is seeking a Safeguards Enforcement Analyst focused on Conventional Weapons to detect and mitigate misuse of our AI systems, applying policy enforcement, threat analysis, and red-team style evals across a range of dangerous technologies.

You will design scalable review workflows, develop evals, and partner with Engineering and Data Science to close gaps, while remaining aware that you may encounter explicit content related to violence or harm.