Product Manager, AI Safety & Safeguards

Anthropic; New York (NY); Hybrid; USD 385,000 - 460,000; Full time

Benefits offered by this job

Office space and amenities

Flexible working hours

Equity donation matching

Generous vacation and parental leave

Job summary

Anthropic is seeking a Product Manager for the Safeguards team to own ideation, design, development and deployment of safeguards systems and related UX. You will work with research and product teams to build detections, evals, interventions, and tools that mitigate deployment and user risks across Claude.ai, 1P API and cloud providers.

You will collaborate with policy, enforcement and engineering to define priorities, articulate tradeoffs, and drive safe, useful AI features in a fast-moving

Qualifications

• Experience coordinating with policy, research, engineering and product teams.
• Ability to design and measure safety-focused product features.
• Strong written and verbal communication of complex concepts.

Responsibilities

• Define safety-by-design features for frontier models across surfaces (Claude.ai, 1P API, cloud).
• Write safety evals and communicate externally about safety.
• Prioritize work with clear technical and business tradeoffs and MVP vs. ideal state.
• Collaborate with policy, enforcement, research and engineering teams.
• Lead development of metrics to assess risk, performance and gaps.

Skills

Tradeoff decisions

User understanding

Cross-functional leadership

Metrics design

Communication

Ambiguity management

Strategy development

Technical translation

Risk assessment

Education

Bachelor's degree

Job description

Anthropic is seeking a Product Manager for the Safeguards team to own ideation, design, development and deployment of safeguards systems and related UX. You will work with research and product teams to build detections, evals, interventions, and tools that mitigate deployment and user risks across Claude.ai, 1P API and cloud providers.

You will collaborate with policy, enforcement and engineering to define priorities, articulate tradeoffs, and drive safe, useful AI features in a fast-moving