Job Description

Opportunity Description

Want to build a career in AI safety research with full financial backing? The ML Alignment & Theory Scholars (MATS) Program – Winter 2027 cohort, run by MATS Research, offers a FULLY FUNDED 12-week research fellowship in Berkeley, California, USA, and London, United Kingdom. The program connects emerging researchers, including US and non-US citizens, with mentors from organizations such as Anthropic, Google DeepMind, OpenAI, Redwood Research, and the Alignment Research Center. Additionally, MATS provides a weekly stipend and covers computing costs, housing, food, and travel. Moreover, the fellowship runs from January 19 to April 10, 2027. Finally, applications for the 2026–2027 cycle are now open.

Key Details

Research Tracks

• Empirical
• Theory
• Strategy and Forecasting
• Policy and Governance
• Systems Security
• Founding and Field-Building
• Biosecurity

Within each track, applicants who advance choose specific “streams,” each led by one or more mentors with a defined research agenda. All backgrounds and experience levels are welcome, and prior AI safety experience is not required.

MATS Winter 2027 Benefits

• Stipend of $1,600 per week, totaling $19,200 for the full 12-week, 40-hour-per-week program. Fellows working fewer hours per week or for fewer weeks receive a prorated stipend.
• Compute a budget of $24,000 per fellow.
• Free housing for the duration of the program.
• Free food, covering lunch and dinner on weekdays.
• Travel to and from Berkeley or London covered.
• J1 visa support for non-US participants who need it to attend in Berkeley full-time.
• Access to a research management team that helps with research strategy, unblocking, and grant applications.
• Option to apply for a funded 6- to 12-month extension after the main program. Roughly 80% of fellows who apply for the extension have been accepted in recent cohorts.

Eligibility Criteria

• Applicants must be 18 years or older before the program start date (January 19, 2027).
• Both US and non-US citizens can apply.
• All backgrounds and levels of experience are welcome.
• Prior AI safety research experience is not required.
• Applicants must be willing to commit 40 hours per week to the program. A reduced commitment of 20 hours per week is possible for exceptionally strong candidates on a case-by-case basis, except for candidates requiring a J1 visa, who cannot participate part-time unless remote or based in the UK.
• Applicants must provide contact details for two references, who are contacted only if the application advances.
• Use of large language models to write any part of the application is not allowed unless a specific work test or form explicitly permits it. MATS states it monitors for this and may disqualify applicants who use LLMs improperly.

How to Apply?

• First, review the tracks and streams on the official Winter 2027 program page and decide which ones fit your research interests.
• Next, complete the Stage 1 general application, which covers your background, short written responses, and two references.
• Then, if you advance, complete Stage 2 central evaluations. Empirical track applicants complete standardized assessments before selecting streams; all other tracks proceed directly to stream selection.
• Afterward, if you advance further, complete Stage 3 mentor review, which may include interviews and work tests.
• Meanwhile, MATS will send offers in early to mid-November 2026, ahead of the program start on January 19, 2027.
• In addition, applicants to the Neel Nanda stream must use a separate form and submit their applications by September 4, 2026.
• Finally, for more information, visit the official Winter 2027 program page.

Apply Now

You may also be eligible:

DSU Toscana Scholarship in Pisa 2026/2027 – FULLY FUNDED

Boston University Free Online Courses 2026 | edX

Mundus MAPP FULLY FUNDED Scholarship 2027–2029