Staff+ Software Engineer – Safeguards

About the role

Anthropic is seeking a Staff+ Software Engineer to join its Safeguards team, building safety and oversight mechanisms for its AI systems. The engineer will help monitor models, detect unwanted behaviors, and prevent disallowed use, ensuring user well‑being and compliance with Anthropic’s terms of service.

Key responsibilities

• Design and implement systems that detect and intervene on unsafe model outputs across all Anthropic products and APIs.
• Develop sandboxed agent architectures and privacy‑preserving data‑access pipelines for safe investigation of abuse signals.
• Build and maintain large‑scale detection stacks, including classifiers, probes, and real‑time intervention mechanisms.
• Collaborate with Trust & Safety teams to translate safety policies into reliable engineering solutions.
• Contribute to the Claude Investigation tool that autonomously reasons over billions of interactions to surface threats.

Required profile

• Staff‑level software engineer with a strong interest in AI safety and trustworthy systems.
• Experience building large‑scale, high‑throughput backend services.
• Ability to work on complex problems at the intersection of security, privacy, and AI.

Required skills