The MATS (ML Alignment & Theory Scholars) Program is a premier research fellowship that connects talented researchers with world-class mentors from organizations like Anthropic, OpenAI, Google DeepMind, and Redwood Research. The program’s primary mission is to reduce frontier AI risk by accelerating technical alignment research and field-building efforts. The fellowship is structured into seven distinct tracks: Empirical, Theory, Strategy & Forecasting, Policy & Governance, Systems Security, Founding & Field-Building, and Biosecurity. Fellows work closely with their mentors on specific research “streams,” often leading to high-impact publications or the development of new safety-focused startups. The program is designed to be an intensive “accelerator” for AI safety careers, providing the resources and community needed to make significant contributions to the field. The application process is rigorous, involving general track evaluations followed by stream-specific work tests and interviews. Once selected, fellows are integrated into a vibrant community of safety researchers, receiving not only financial support but also technical mentorship, compute resources, and a dedicated research management team to help them navigate the complexities of AI alignment. Target Audience: Researchers, engineers, and field-builders with a strong technical or policy background who are committed to working on AI safety and alignment. Geographic Eligibility: In-person (Berkeley, CA or London, UK) or remote, depending on the specific mentor’s requirements. Tangible Benefits: Competitive stipend, compute credits, housing and meals (for in-person cohorts), research management support, and access to an elite network of AI safety professionals. Application Deadline: June 7, 2026 Program Timeline: Program starts September 28, 2026