.webp)
The Residency exists to help each Resident maximize the impact of their research. To do this, every project is held to the same rigorous standards of evidence and review. Projects are continually assessed on their progress to determine the type of support needed, whether it be continued research support, new collaborations, or longer-term opportunities. Because Residents progress at different rates, our consistent evaluation standard ensures that each Resident receives targeted, personalized support that enables them to do their best work.
Each application goes through a four step review process.
In the past 4.5 years, we have helped produce more than 200 research publications with over 12,000 collective citations; our organizational h-index is 50.
MATS fellows have helped develop new research agendas, including sparse auto-encoders for AI interpretability, activation/representation engineering, emergent misalignment, inoculation prompting, developmental interpretability, computational mechanics, glitch token analysis, evaluating situational awareness, gradient routing, externalized reasoning oversight, conditioning predictive models, formalizing natural abstractions, and more!
10% of alumni have co-founded AI safety organizations or research teams during or after MATS.
MATS alumni-founded organizations include Aether, AI Safety Argentina, Algoverse AI Safety Fellowship, Apollo Research, ARENA, Athena, Atla, Cadenza Labs, Catalyze Impact, Center for AI Policy, Contramont Research, Coordinal Research, Decode Research, Freestyle Research, Fulcrum, General Analysis, Groundless, Leap Labs, LISA, Luthien Research, MIRI Technical Governance Team, Poseidon Research, Principled Agents, PRISM Eval, Simplex, SL5 Taskforce, StakeOut AI, Timaeus, Theorem Labs, Watertight AI, WeaveMind, and Workshop Labs.
80% of alumni are now working in AI alignment, transparency, and security.
MATS alumni have been hired by leading organizations like Anthropic, Google DeepMind, OpenAI, Meta AI, UK AISI, Redwood Research, METR, RAND CAST, Coefficient Giving, ARC, FAR.AI, Apollo Research, Truthful AI, Goodfire, LawZero, MIRI, CAIF, Center on Long-Term Risk, Beneficial AI Foundation, SaferAI, Haize Labs, EleutherAI, Harmony Intelligence, Conjecture, and joined academic research groups like UC Berkeley CHAI, NYU ARG, NU Bau Lab, Mila, and MIT Tegmark Group.
For experienced academics, researchers, and industry professionals from adjacent fields (e.g. ML, policy, governance) seeking opportunities in AI safety.
You may already have the expertise to make important contributions to AI safety but need a dedicated runway to build a portfolio that will catch the attention of employers, collaborators, and experts in the field. Some examples: A professor with a strong record in empirical machine learning may need to demonstrate how their work connects to safety-relevant questions. A MATS alumnus may be temporarily unavailable to join a frontier lab because of visa timelines, hiring cycles, or limited team capacity. A senior industry researcher may have extensive technical expertise but lack published AI safety research that hiring committees can evaluate. A policy or governance professional may bring deep expertise in regulation or national security but need to build a portfolio that translates their perspective into concrete AI safety research.
The Bridge track fills this gap by helping experienced researchers establish a credible AI safety research portfolio. Residents pursue an independent research agenda while producing reviewable work that demonstrates their expertise and relevance. Through the Placement Desk, MATS also provides career development support, tracks relevant opportunities, facilitates introductions, and provides referral letters where appropriate.
For researchers new to AI safety, the track begins with a Safety Orientation Week, followed by a three-week Translation Sprint, where participants take a piece of prior work and reframe it around a safety-relevant threat model. This helps them establish an initial research direction while connecting their existing expertise to concrete safety questions.
The Bridge track also supports more substantial career transitions. Established academics may use a supported sabbatical, fellowship, or postdoctoral placement of up to two years to redirect their research toward AI safety, field-building, or founding a new organization. While normally taken as full-time leave, part-time arrangements may be considered on a case-by-case basis. This reduces the financial and reputational risks associated with a mid-career transition.
For recent PhD graduates, the Bridge track offers an alternative to a conventional postdoctoral position. Participants pursue an AI safety research agenda of their choice, work with senior researchers from frontier labs, and produce reviewable outputs that support subsequent placement into research organizations.
For applicants who want a long-term staff researcher role rather than a PhD or postdoc, and who need an institutional setting in which to develop an independent research agenda.
Many capable researchers who do not fit a traditional academic trajectory have difficulty finding long-term roles that match their depth and independence. The Anchor track provides an alternative research pathway for engineers, evaluators, and technical or policy practitioners whose work is fundamentally research. Residents are held to the same rigorous standards of evidence, research quality, and demonstrated progress expected across the Residency.
During the first four weeks, Residents develop a Research Agenda Proposal. It is reviewed by the MATS Residency team with input from the program's advisors where appropriate, and Residents receive feedback within the first six weeks. Beginning in Block 2, every Resident is paired with a peer accountability partner from the active Residency community for check-ins every two weeks.
At the 12-month milestone, Residents demonstrating strong progress enter the Independence Pathway Review: a formal assessment of portfolio quality, research judgment, execution, and overall fit, informed by external reviewers. The review provides substantive feedback to guide the next 12 months of research.
For researchers founding new AI safety organizations or research programs, who need structural support to move from idea to execution.
Apollo Research and Timaeus (now Resolution) both emerged from MATS. The Incubation track is designed to make outcomes like these more repeatable by providing dedicated support for founders building new AI safety organizations. Residents receive structured proof-of-concept support, external review from people with founding and funder experience, and guidance on governance scaffolding from Day 1. We help founders expand their professional network, establish credibility with collaborators and funders, and access the relationships and resources needed to launch a successful organization.
The track uses a rigorous, committee-based evaluation process to maintain a high bar and ensure that the most promising and feasible ideas are championed. It provides structured support from early idea development through validation, organizational design, and fundraising.
At the end of Block 1, each team takes part in a Funding Evaluation Workshop. The panel includes two external reviewers with experience founding AI safety organizations and one reviewer with funding or evaluation expertise. The panel's input is calibrated to what the team has actually built and learned, rather than the strength of the original proposal alone.
All Incubation Residents also submit a Governance and Publication Plan at admission. This covers the proposed legal structure, oversight arrangements, publication policy, and the organization’s approach to AI safety risks such as information hazards and dual-use research.
Applicants may apply as an individual if they do not have a founding team, or a founding team can submit one application collectively. Since the Residency focuses on amplifying AI safety research, at least one person on the application must demonstrate safety-relevant research experience or evidence of research capability. Additional co-founders in operating or non-research roles are welcome as part of the application and are not expected to meet the research threshold themselves.
The quickest way to choose is by where you want to end up.

Bridge places you into a role elsewhere, with the Placement Desk and referral letters. Anchor is itself the long-term home, a staff researcher path you stay in. Incubation ends in a new organization you have launched.
Two cases are worth calling out. If you are transitioning into AI safety but want a lasting research home rather than to be placed into an existing organization, Anchor, not Bridge, is likely your track. And if your direction could be pursued either as your own research or built into an organization, ask whether the work needs an institution to exist: if you mainly need time, support, and independence to do the research yourself, choose Anchor; if it only succeeds as a new organization with a team, funding, and its own governance, choose Incubation. For example, a new interpretability agenda you would pursue as a researcher is Anchor; a new independent evaluations organization that needs staff, infrastructure, and funders to deliver its mission is Incubation.