MATS AI Safety Fellowship Autumn 2026

Applications are now open for the MATS (Machine Learning Alignment & Theory Scholars) Autumn 2026 Fellowship, one of the world’s leading AI safety research programs. The fellowship connects talented researchers, engineers, policymakers, and field-builders with world-class mentors working on artificial intelligence alignment, governance, security, forecasting, and emerging AI challenges.

The Autumn 2026 cohort will run for 10 weeks from September 28 to December 4, 2026, with activities taking place in Berkeley, California, USA, and London, United Kingdom. Fellows will work alongside mentors affiliated with leading AI organizations and research institutions, including organizations such as Anthropic, Google DeepMind, OpenAI, ARC, and Redwood Research.

Fellowship Overview

• Program: MATS AI Safety Fellowship
• Cohort: Autumn 2026
• Duration: 10 Weeks
• Dates: September 28 – December 4, 2026
• Locations: Berkeley (USA) and London (UK)
• Funding: Fully Funded
• Application Deadline: June 7, 2026
• Extension Opportunity: Selected fellows may receive a 6–12 month funded extension after the main program.

About the MATS Program

The MATS Program is an independent research fellowship designed to train the next generation of researchers working on AI alignment, interpretability, governance, security, and other areas related to safe and beneficial artificial intelligence. Fellows receive intensive mentorship, participate in workshops and seminars, collaborate with peers, and contribute to impactful research projects.

The program aims to help talented individuals transition into long-term careers focused on reducing risks associated with advanced AI systems while advancing research that improves transparency, safety, and governance.

Research Tracks

Applicants can apply to one or more of several specialized research tracks:

Empirical AI Safety

Focuses on experimental and practical approaches to AI alignment, interpretability, evaluations, and safety mechanisms.

Theory

Research on foundational questions related to AI alignment, agent behavior, decision theory, and long-term AI safety.

Policy and Governance

Examines policy frameworks, international cooperation, regulation, and governance strategies for advanced AI systems.

Strategy and Forecasting

Focuses on forecasting AI development, geopolitical implications, scenario planning, institutional responses, and long-term strategic questions surrounding advanced AI.

Systems Security

Research on AI system robustness, cybersecurity, adversarial risks, and safety safeguards.

Biosecurity

A newly introduced track exploring the intersection of advanced AI and biological risks, including AI-enabled biosecurity challenges and mitigation strategies.

Founding and Field-Building

Designed for entrepreneurs, operators, ecosystem builders, and leaders interested in launching impactful AI safety organizations and initiatives.

Fellowship Benefits

Selected fellows receive:

• Mentorship from leading AI researchers and experts
• Access to research workshops and seminars
• Participation in a global AI safety community
• Networking opportunities with researchers and industry leaders
• Professional research support
• Potential funded extension opportunities lasting 6–12 months
• Access to research hubs in Berkeley and London
• Collaboration opportunities with leading AI safety organizations.

Who Can Apply?

The fellowship welcomes applicants from a wide variety of backgrounds, including:

• Artificial Intelligence
• Machine Learning
• Computer Science
• Mathematics
• Physics
• Economics
• Public Policy
• Cognitive Science
• Biology
• Public Health
• Entrepreneurship
• Research Management
• Governance and Strategy

Applicants do not necessarily need prior AI safety experience, although strong motivation, research potential, technical aptitude, or relevant professional experience are important selection factors.

Application Process

The application process consists of multiple stages:

Stage 1: General Application

Applicants complete a general application and select research tracks that align with their interests. The application typically takes 1–2 hours to complete.

Stage 2: Additional Evaluations

Depending on the selected track, applicants may complete:

• Coding assessments
• Research proposals
• Written evaluations
• Work tests
• Additional questionnaires
• Track-specific assessments.

Stage 3: Mentor Selection and Interviews

Shortlisted applicants interview with potential mentors to determine mutual research fit and project alignment. Final fellowship offers are then extended to successful candidates.

Why Apply?

The MATS Fellowship is widely regarded as one of the most influential entry points into the AI safety ecosystem. Fellows gain direct mentorship from leading experts, access to cutting-edge research communities, and opportunities to contribute to some of the most important questions surrounding the future of artificial intelligence. Community discussions frequently describe the program as highly competitive and well-regarded within the AI safety field.

For aspiring AI researchers, policymakers, strategists, and founders, the fellowship offers a unique platform to build expertise, expand professional networks, and make meaningful contributions to the future of AI.

Application Deadline

June 7, 2026

Opportunity Summary

• Host: MATS Program
• Fellowship: Autumn 2026 AI Safety Fellowship
• Duration: 10 Weeks
• Locations: Berkeley (USA) & London (UK)
• Funding: Fully Funded
• Fields: AI Safety, Governance, Security, Forecasting, Biosecurity, Entrepreneurship
• Deadline: June 7, 2026
• Extension Opportunity: 6–12 Months for Selected Fellows