CBAI Fall Research Fellowship in AI Safety 2026

CBAI Fall Research Fellowship in AI Safety 2026 is sponsored by Cambridge Boston Alignment Initiative (CBAI). 10-week research fellowship in Cambridge, MA, focused on AI safety topics including evaluations, safeguards, threat modeling, and governance. Part of a larger cohort with office space and community support.

AI Safety Research Fellowship — Cambridge Boston Alignment Initiative CBAI Fall Research Fellowship in AI Safety ‘26 The Cambridge Boston Alignment Initiative will be running our next AI Safety Research Fellowship from October 13 to December 18. This is an intensive, fully-funded, ten-week research program hosted in Cambridge, Massachusetts.

It is designed to support talented researchers aiming to advance their careers in AI safety, covering interpretability, multi-agent safety, formal verification, risk management frameworks, and other technical and governance domains.

Fellows work closely with established mentors and in-house research managers, participate in weekly workshops and speaker events, and gain invaluable research experience and networking opportunities within the vibrant AI safety community at institutions like Harvard, MIT, Northeastern, Boston, and leading AI safety organizations.

Our inaugural fellowship cohort members have joined Goodfire and Redwood, established their own research group, been accepted into NeurIPS and ICLR, and shared their reports with policymakers in DC. Please fill out this form to stay updated about our upcoming cohorts. We will be able to sponsor J-1 visas effective for the Spring cycle, which will be released in late September.

The application deadline is September 6, 11:59 pm Eastern Time. We host a speaker event series under the fellowship program. $15,000 over the ten-week fellowship.

For candidates needing relocation, we will provide housing through Airbnb and short-term corporate rentals. Free weekday lunches and dinners, plus snacks, coffee, and beverages. 24/7 office access in Harvard Square, a few minutes from Harvard Yard.

Weekly individual mentorship (1–2 hrs/week) from researchers at renowned institutions such as RESI, Harvard, MIT, Northeastern, Goodfire, AVERI, METR, Robocurve, and others. Dedicated in-house research managers provide high-touch research support to strengthen your research skills, clarify the research direction, and contribute to your career trajectory in AI safety research.

Weekly speaker event series with renowned researchers across AI safety and biosecurity, and events, workshops, and socials with both communities in the city. We will provide ~$1,500 per week of compute support to each fellow in the form of API credits and on-demand GPUs. However, we expect this amount to be flexible for empirical projects with intensive compute needs.

In addition, we cover the costs associated with a conference, should your paper be accepted to a reputable conference. On a case-by-case basis, we may be able to provide extension funding support for up to 6 months. Historically, we’ve supported more than 75% of extension requests.

Assistant Professor, Northeastern University Member of Technical Staff, Goodfire AI Assistant Professor, Harvard Kennedy School AI Science Advisor, California Council on Science & Technology CEO and Co-founder, Robocurve Member of Technical Staff, Goodfire AI Principal Investigator, Praxis; Assistant Professor, GWU Researcher, Institute for Responsible Superintelligence (RESI) Member of Technical Staff, METR CSO, Institute for Responsible Superintelligence (RESI) Postdoctoral Researcher, Northeastern University Co-Founder, AVERI; Founder, AI Incident Database Gordon McKay Professor of Computer Science, Harvard University Senior Researcher, MIT AI Risk Initiative Research Fellow, Harvard University Director of Standards, AVERI Co-founder, Simplex; Research Lead, Astera Institute Project Director, MIT AI Risk Initiative Associate Professor of Law, University of Houston; Founder and Executive Co-Director, Center for Law & AI Risk Cofounder, Simplex; Colead, Astera Research Scientist, MIT AI Risk Initiative Group Leader, Physics of Intelligence Group at Harvard University Assistant Professor, University of Houston; Non-Resident Senior Fellow, Institute for Law & AI Professor of International Law, Harvard Law School We welcome applications from anyone deeply committed to advancing the safety and responsible governance of artificial intelligence.

Ideal candidates include: Undergraduate, Master's, and PhD students — and Postdocs — looking to explore or deepen their engagement in AI safety research. Recent graduates who are aiming to transition into technical or governance AI safety research. Individuals who are passionate about addressing the risks associated with advanced AI systems.

We think the most promising undergraduates are severely neglected in the field and can achieve great things. We will put a heavy emphasis on undergraduate candidates, as some mentors will have undergraduate quotas, and we encourage undergraduates to apply.

Eligibility: CBAI does not employ fellows, but all fellows should understand that the stipend is characterized as taxable income, so fellows should have the legal right to work in the United States. Previous fellows who are on F-1 or J-1 visas have had success using OPT and CPT for their fellowship participation. CBAI staff is not able to provide tax or legal advice to fellows.

We are unable to sponsor visas for this program. Referral: If you refer a successful fellow to us, you can get a $500 Amazon gift card! Refer fellows here !

Note: Participants must be 18 years or older to apply for this fellowship. Our application process consists of four steps: 15-minute interview with CBAI Mentor-specific question, test task, or code screening (if applicable) An interview with the mentor Applications are due September 6 at 11:59 pm.

Research focused on reducing catastrophic risks from advanced AI, including alignment strategies, interpretability, robustness, formal verification, multi-agent safety, and scalable oversight. Research and analysis aimed at shaping how AI is developed, deployed, and governed, including policy frameworks, regulatory approaches, international coordination, and institutional mechanisms for managing risks from powerful AI systems.

Research at the intersection of technical AI safety and policy, including compute governance, model evaluations, standards development, and institutional mechanisms for ensuring advanced AI systems are safe and beneficial.

Economics of Transformative AI Research on the economic causes and consequences of advanced AI development, spanning empirical and modeling approaches to growth, productivity, and labor market dynamics, as well as policy questions around industrial strategy, market concentration, and the political economy of AI governance. For more information and your questions, please reach out to fellowships@cbai. ai.

Frequently Asked Questions I'm a prospective applicant from overseas. Can I secure my own visa and join the fellowship? Unfortunately, no. Per the US laws, we cannot have participants with no work permit join our fellowship program.

If you don’t have a work permit (e.g., permanent resident, H-1B visa holder, or a student with OPT/CPT), we cannot have you in the program. What will a typical day look like during the fellowship? Your day will typically begin at our co-working space located in Harvard Square, where you’ll dive into your research project.

Around lunchtime, you will attend the weekly speaker event series with virtual or in-person guests. Afternoons usually involve dedicated deep-work sessions focused on your research projects. In the evenings, fellows frequently gather for dinner, followed by various group activities, ranging from paper reading groups to board game nights.

Throughout the week, you'll also have scheduled one-on-one check-ins with your research manager, members of the CBAI team, and coffee chats with other fellows, as well as numerous opportunities to engage with other fellowship events. What outcomes can I expect from participating in the fellowship? Fellows are expected to produce written outputs suitable for sharing publicly.

These could be in the form of blog posts, interactive demos (if applicable), pre-prints, and paper submissions to ICLR, ICML, NeurIPS, and workshops in these conferences.

While there are no rigid guidelines concerning the format or scope of these outputs, we will actively support fellows by identifying relevant conference opportunities, such as NeurIPS, and assisting them in showcasing their research to broader AI safety and academic communities.

We anticipate this fellowship will lead to ongoing mentorship relationships and foster new collaborations among fellows, providing a foundation for continued engagement and future joint projects. What type of events will you organize?

We will organize a speaker series with renowned researchers and peer researchers in the Cambridge area and virtual speakers to expose you to frontier research and other relevant research areas that you might be interested in pursuing post-fellowship. We will also organize a range of activities, including networking events with researchers in the CBAI network and researchers in the area, research workshops, and wargames.

Every week, we will host socials and outings to the city as well as board game nights and other activities. Additionally, you'll have opportunities to interact regularly with AI safety researchers from institutions such as Harvard and MIT, through attending the co-hosted events with research groups in those universities.

Additionally, if you’re working with a mentor in a lab, you’re expected to use their facilities 1-2 times a week to expose yourself to other research and discussions in the group. This summer, CBAI will also organize an AIxBiosecurity Research Fellowship Program, with approximately 45 participants in both programs, creating a dynamic environment with frequent activities & deep discussions.

Through MAIA and AISST , AI safety student groups that are fiscally sponsored and operated by CBAI, we will host networking events and socials for junior researchers. How large will be the cohort? We’ll accept 30 fellows for our AI safety research fellowship program, and 15 fellows for our AIxBiosecurity research fellowship program, which will be released in mid-April.

Although two programs are technically separate, our offices are two minutes from each other and fellows will stay in the same building and attend mostly the same speaker events and socials frequently. Is CBAI officially affiliated with Harvard or MIT? While CBAI fiscally sponsors and operates the AI Safety Student Team at Harvard and MIT AI Alignment , we are not officially affiliated with Harvard or MIT.

However, we work closely with research groups and labs at both universities, and others, including Northeastern and Boston, and fellows working with mentors from labs in these universities may receive affiliate cards to get access to the premises. Note: If you are a student at a university affiliated with BorrowDirect , you are eligible for a library card granting access to the libraries at Harvard and MIT.

What does housing support entail? For those joining us outside the Cambridge-Boston area, we will provide housing through Harvard dorms located a few minutes away from our offices. If you prefer to arrange your own housing, please contact us; we may be able to provide housing reimbursement.

Can I still be accepted into the program if none of the mentors choose me? Mentor matching is a key component of our program, ensuring each fellow receives dedicated support and guidance. However, in rare cases, we might extend a conditional offer to candidates after the interview round.

This offer would be conditional upon either receiving an offer from one of the mentors listed on our website or an external mentor that fellowship organizers agree on. Is it possible to attend only part of the program in person? Our default expectation is that all fellows participate in person for the full duration of the fellowship.

However, we understand there may be special circumstances, and we are open to evaluating these on a case-by-case basis. How can I increase my chances of being accepted into the program? Successful applicants typically demonstrate clear alignment with the goals and values of our program: a well-articulated interest in AI safety, strong motivation for pursuing research in this area, and a very clear career plan for the post-fellowship.

To strengthen your application, clearly outline your relevant background , skills , and past experiences , explain specifically why you're interested in this fellowship , and highlight the impact you hope to achieve through your participation, as well as how you will utilize this fellowship for your post-fellowship. What's the difference between my research manager and my mentor?

Your research manager provides ongoing support with project management and progress tracking, and may offer research guidance depending on their background. Your mentor is an external expert matched to your specific research area, offering specialized domain knowledge, targeted feedback, and substantive input on your research direction. Can I join virtually or part-time?

Barring exceptional cases (e.g., someone with a disability that would make attending in-person extremely difficult for them), we will not accept virtual candidates. We do not allow part-time participation.