Anthropic Fellows Program for AI safety research: applications open for May & July 2026
Anthropic's Fellows Program offers mentored, empirical AI safety research opportunities for engineers and researchers, focusing on interpretability, robustness, and public research outputs.
Funder: Anthropic
Due Dates: May 2026 (Cohort start) | July 2026 (Cohort start) | Rolling (applications reviewed as received)
Funding Amounts: $3,850 USD/week (or equivalent in GBP/CAD) for 4 months; ~$15,000/month compute/research expenses
Summary: Four-month, mentored empirical AI safety research fellowships for engineers and researchers, with funding and public research outputs.
Key Information: Full-time work authorization and physical presence in the US, UK, or Canada required; no visa sponsorship.
Description
The Anthropic Fellows Program offers a unique opportunity for engineers and researchers to conduct high-impact, empirical research on AI safety, security, interpretability, and related topics. Fellows work full-time for four months, collaborating closely with Anthropic mentors on projects aligned with the organization's research priorities. The program emphasizes producing public outputs such as research papers, tools, or datasets, and provides direct mentorship, funding, and access to compute resources. Research areas include scalable oversight, adversarial robustness, model organisms of misalignment, mechanistic interpretability, AI security, model welfare, and more. The program is designed to support both early-career and mid-career technical talent seeking to transition into AI safety research.
