AI Safety

MATS Autumn 2026 Fellowship opens with new Biosecurity & Founding tracks

10-week fully-funded AI safety fellowship with $5K/month stipend plus compute.

Deep Dive

MATS (ML Alignment & Theory Scholars) Research is now accepting applications for its Autumn 2026 cohort, the first fall program in a new three-cohort-per-year schedule. The 10-week residential fellowship runs from September 28 to December 5, 2026, in Berkeley, California and London, UK. Fellows receive a $5,000/month stipend, an $8,000/month compute budget, plus housing, meals, and travel covered. Research tracks include Empirical Theory, Strategy & Forecasting, Policy & Governance, Systems Security, and two brand-new options: Biosecurity (focused on preventing catastrophic biological risk from AI) and Founding & Field-Building (for high-agency generalists launching AI safety startups). Mentors come from leading labs like Anthropic, Google DeepMind, OpenAI, Redwood Research, and the AI Futures Project. J1 visa support is available, and over 80% of participants get a 6-to-12 month extension with a higher $7,680/month stipend while retaining the compute budget.

MATS has produced 527 alumni, over 100 mentors, and more than 200 publications with 12,300+ citations. Nearly 80% of pre-2025 graduates are working in AI safety/security, and 10% have co-founded active AI safety startups. The program explicitly seeks talent from non-traditional backgrounds—including technical researchers without prior ML experience—if they demonstrate strong reasoning and research potential. With the shift to three cohorts per year, MATS aims to expand the AI safety talent pipeline urgently. Applications close June 7, 2026 (AoE) at matsprogram.org/apply.

Key Points
  • $5,000/month stipend + $8,000/month compute budget; housing/meals/travel covered
  • New tracks: Biosecurity and Founding & Field-Building, plus six existing tracks
  • Over 80% of fellows receive a 6–12 month extension with $7,680/month stipend; apply by June 7, 2026

Why It Matters

Funded AI safety fellowship trains the next generation to tackle catastrophic risks from advanced AI.