SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 245,000 - 285,000 / annual
Anthropic is seeking a Policy Design Manager to own the company's conventional weapons policy within its Safeguards organization. This role focuses on defining and enforcing boundaries around how Claude AI can be used in weapons-related contexts, balancing legitimate research and engineering with misuse prevention.
The core challenge is navigating dual-use technology: the same capabilities that enable civilian engineering and research can contribute to weapons systems. You will develop threat models, build evaluations, and create detection systems that operationalize the distinction between prohibited weapons development and legitimate work.
Key responsibilities include:
- Owning and maintaining Anthropic's conventional weapons policy, defining what activities the models should and should not support
- Building threat models and evaluations to measure how Claude could contribute to weapons development, including software and autonomy components
- Partnering with engineering teams to translate policy into model guardrails, detection systems, and enforcement tooling
- Serving as the subject-matter expert for escalations involving conventional weapons content and rapid response to emerging risks
- Communicating policy reasoning across product, engineering, legal, and leadership to both technical and non-technical audiences
- Engaging external experts, government, and industry partners to strengthen policy and enforcement
You will need deep, applied expertise in weapons systems and the ability to translate complex technical evidence into sound policy judgments. Experience in service research laboratories, defense research agencies, or weapons design/manufacturing is essential. You must write clear, operationally precise policy and explain complex topics to non-specialists. Understanding of legal frameworks governing weapons and their transfer across jurisdictions is required, along with the ability to conduct rigorous technical analysis using open sources.
Preferred qualifications include working knowledge of machine learning and LLM fundamentals, hands-on engineering experience in weapons-relevant domains (systems engineering, robotics, autonomy, guidance/navigation/control, sensors, aerospace, mechanical engineering, materials, embedded software), experience with arms export controls (ITAR/EAR), experience building or evaluating classifiers including LLM-based ones, and trust & safety or product policy experience at technology platforms.
