Archived 🎓 Fellowships

MATS Autumn 2026 — AI Safety Research Fellowship

A fully-funded 10-week AI safety research fellowship (Sep 28–Dec 4, 2026) in Berkeley or London. Fellows work with mentors from Anthropic, OpenAI, Google DeepMind, Redwood Research, and ARC. Over 80% of alumni continue into AI safety roles. New tracks: Founding & Field-Building and Biosecurity.

fellowshipAI-safetyalignmentinterpretabilitygovernanceglobalfully-fundedBerkeleyLondon

Why it matters

One of the most credentialed AI safety fellowships globally — mentors include Anthropic, DeepMind, and OpenAI researchers, and over 80% of alumni go on to full-time AI safety roles. New Founding track makes it relevant for field-builders, not just researchers.

Provider 🏛 MATS Research (Machine Learning Alignment & Theory Scholars)

Deadline 📅 June 7, 2026

Value 💵 $5,000/month stipend + $8,000/month compute + housing & meals

Region 🌍 Global (in-person: Berkeley CA or London UK)

Eligibility Open globally. Requires strong motivation to work on AI safety and evidence of technical aptitude or research potential. Prior AI safety experience helpful but not required. Tracks available in alignment, interpretability, governance, security, biosecurity, and founding/field-building.

Best For 👤 Researchers, Students

Effort Level 🟡 Medium

Published Jun 7, 2026

Category 🎓 Fellowships

Last Verified ✓ Jun 7, 2026

Source ✦ Official Source ↗

MATS (Machine Learning Alignment & Theory Scholars) is an independent research fellowship connecting talented researchers with top mentors working on AI alignment, transparency, and security. The Autumn 2026 cohort is MATS’s first autumn run — part of a shift to three cohorts per year.

Fellowship details:

Duration: 10 weeks, September 28 – December 4, 2026
Locations: Berkeley, CA (USA) and London, UK (mentor-dependent)
Cohort: Selective — hundreds of applicants per cycle

What fellows receive:

$5,000/month stipend + $8,000/month compute budget
Office space, housing, and meals
J-1 visa support for Berkeley office
Mentorship from Anthropic, OpenAI, Google DeepMind, Redwood Research, ARC, AI Futures Project, and more
Talks, workshops, and a global alumni network
6–12 month funded extension for top fellows (80%+ of alumni)

Research tracks (Autumn 2026):

Technical Alignment
Interpretability
Governance & Policy
Systems Security
Biosecurity (new)
Founding & Field-Building (new)

Key dates:

Application deadline: June 7, 2026 (AoE — today, final hours)
Final offers: Late July / early August 2026
Programme start: September 28, 2026

Apply at matsprogram.org/apply. Note: MATS accepts applicants from all countries, including the Global South.

← More Fellowships ← All Opportunities