Product Manager, AI Safeguards & Safe Frontiers

Anthropic Limited; San Francisco (CA); Híbrido; USD 305.000 - 385.000; Jornada completa

Descripción de la vacante

Anthropic Limited is seeking a Product Manager for the Safeguards team to own the ideation, design, and deployment of safeguards systems and related UX. You will collaborate with research and product teams to build detections, evals, interventions, and tools to mitigate deployment risks across cloud platforms.

You will work in a fast-paced, ambiguous environment to advance frontier models safely for users, with a focus on measurable impact and clear communication to diverse stakeholders.

Formación

• Ability to design and compare safety-related options across policy, research, and engineering teams.
• Strong understanding of AI risk, safeguards, and responsible deployment in cloud environments.
• Experience creating measurable safety metrics and interpreting results.
• Proven track record aligning product strategy with safety requirements in ambiguous spaces.

Responsabilidades

• Define safety-by-design approaches upstream and defenses downstream for frontier models.
• Write safety evals and communicate safety concepts to external stakeholders.
• Prioritize impact with clear problem definition, tradeoffs, and MVP vs. ideal state.
• Coordinate with policy, enforcement, research, and engineering teams across functions.
• Develop metrics to monitor area performance and inform future planning.

Conocimientos

Technical tradeoffs

Safeguards expertise

Cross-functional leadership

Metrics for risk

Clear communication

Educación

Bachelor's degree or equiv

Descripción del empleo

Anthropic Limited is seeking a Product Manager for the Safeguards team to own the ideation, design, and deployment of safeguards systems and related UX. You will collaborate with research and product teams to build detections, evals, interventions, and tools to mitigate deployment risks across cloud platforms.

You will work in a fast-paced, ambiguous environment to advance frontier models safely for users, with a focus on measurable impact and clear communication to diverse stakeholders.