Software Engineer, Safeguards

About this role

Anthropic is seeking a Software Engineer for their Safeguards team to build safety and oversight systems for AI models. You'll develop monitoring tools, detection mechanisms, and multi-layered defenses to prevent misuse and ensure user protection across their API platforms.

What you'll do

• Develop monitoring systems to detect unwanted API partner behaviors and create automated enforcement actions
• Build abuse detection mechanisms and infrastructure for AI systems
• Surface abuse patterns to research teams to improve model training and hardening
• Create robust, real-time, scalable safety defense systems
• Surface findings in internal dashboards for analyst review

What they're looking for

• Python and TypeScript proficiency
• Full-stack software development
• Abuse and fraud detection systems
• Technical communication with non-technical stakeholders
• AI/ML trust and safety mechanisms
• Prompt engineering and adversarial input understanding
• Internal tooling and operational systems

Benefits

• Annual salary: $320,000–$485,000 USD
• Hybrid work policy with 25% minimum office time
• Visa sponsorship available
• Work on AI safety and beneficial AI systems
• Collaborative team of researchers and engineers