MATS Program Autumn 2026: A Deep Dive for Aspiring AI Safety Researchers
The field of Artificial Intelligence (AI) is rapidly advancing, and with this progress comes a growing need for individuals dedicated to ensuring its safe and responsible development. The MATS Program, a 10 to 12-week research fellowship, offers a unique opportunity for emerging researchers and field-builders to contribute to AI alignment, interpretability, governance, and security. This program aims to provide the necessary structure, resources, and mentorship to foster impactful research and launch careers in AI safety. The Autumn 2026 cohort is now accepting applications, presenting a chance to collaborate with leading experts and work on critical AI risk reduction efforts.
Understanding the MATS Program’s Mission
The MATS Program is designed to cultivate the next generation of AI safety experts. It focuses on practical research and field-building initiatives aimed at mitigating risks associated with advanced AI systems. By connecting fellows with mentors from top organizations like Anthropic, OpenAI, and Google DeepMind, the program ensures access to cutting-edge knowledge and guidance. The fellowship provides a supportive environment where participants can concentrate on producing significant work that advances the safety and reliability of AI.
Key Benefits for Fellows
Participants in the MATS Program can expect substantial support to facilitate their research endeavors. The program offers a generous stipend of $12.5k, along with $20k in compute resources, enabling fellows to tackle complex computational challenges. Additionally, fellows receive free housing and meals, removing common financial barriers to intensive research. Travel expenses are also covered, and J1 visas are provided if needed, making the program accessible to international applicants. This comprehensive support system allows fellows to dedicate their full attention to their research projects.
Eligibility and Application Requirements
The MATS Program welcomes applicants from a wide array of academic and professional backgrounds. Whether your expertise lies in machine learning, mathematics, computer science, policy, economics, physics, or cognitive science, you are encouraged to apply. The core requirements are a strong commitment to AI safety and demonstrated technical aptitude or research potential. While prior experience in AI safety is beneficial, it is not mandatory. The application process involves a basic profile submission, an optional record of publications, and responses to two short questions. Some specialized tracks may require additional responses.
The Application Process in Detail
To apply for the MATS Program Autumn 2026, prospective fellows must complete an online application form. This form gathers essential information about the applicant’s background and motivations. Applicants will be asked to provide basic profile details and can optionally include a list of their published works. Two short response questions are designed to gauge the applicant’s understanding of AI safety challenges and their approach to research. Applicants will also select their preferred research tracks, and some tracks may have specific additional questions to ensure a good fit between the fellow and the research area.
Frequently Asked Questions
What is the MATS Program?
The MATS Program is a 10 to 12-week research fellowship designed to help aspiring researchers contribute to AI safety and build careers in the field.
Who is eligible to apply for the MATS Program?
Anyone with a strong commitment to AI safety and demonstrated technical aptitude or research potential is encouraged to apply, regardless of their specific academic background.
What kind of support does the MATS Program offer fellows?
The program provides a stipend, compute resources, free housing and meals, and covers travel expenses, with J1 visas available for international applicants.
How do I apply for the MATS Program?
You need to complete an online application form, providing your basic profile details, optionally listing publications, and answering two short questions about AI safety and your research approach.
