Anthropic Fellows Program: AI safety and security research

A four-month, full-time funded fellowship for empirical AI safety and security research with mentorship, usually ending in a public paper. No PhD is required, but strong Python and a quantitative background are expected. Applicants need existing work authorization in the US, UK or Canada.

• Typical window: rolling, with several cohorts a year
• Stipend and terms may change between cohorts

Official page: alignment.anthropic.com/2025/anthropic-fellows-program-2026/