CBAI Fall AI Safety Research Fellowship 2026: Fully Funded Programme With a $15,000 Stipend, Housing and Compute Support

The Cambridge Boston Alignment Initiative is offering an intensive, fully funded AI Safety Research Fellowship designed to help promising researchers develop practical experience, produce meaningful research and build careers focused on the safe and responsible development of artificial intelligence.

The CBAI Fall Research Fellowship in AI Safety 2026 will run for ten weeks in Cambridge, Massachusetts, from October 13 to December 18, 2026. Selected participants will receive a US$15,000 stipend, furnished accommodation, weekday meals, dedicated research space, individual mentorship, professional research support and substantial computing resources.

The programme covers technical and policy-oriented fields, including AI alignment, interpretability, formal verification, multi-agent safety, AI governance, technical governance and the economics of transformative artificial intelligence.

Applications for the Fall 2026 cohort closed on September 6, 2026, at 11:59 PM Eastern Time. No official deadline extension has been announced. Individuals interested in future opportunities can register for updates because CBAI expects to release information about its Spring cycle in late September.

CBAI Fall AI Safety Research Fellowship 2026 at a Glance

• Organiser: Cambridge Boston Alignment Initiative
• Programme: Fall Research Fellowship in AI Safety 2026
• Location: Cambridge, Massachusetts, United States
• Programme format: Primarily full-time and in person
• Duration: Ten weeks
• Programme dates: October 13–December 18, 2026
• Stipend: US$15,000
• Accommodation: Provided for participants relocating to Cambridge
• Meals: Free weekday lunches and dinners, snacks, coffee and beverages
• Workspace: Twenty-four-hour access to an office in Harvard Square
• Mentorship: Approximately one to two hours of individual mentorship each week
• Compute support: Approximately US$1,500 per week through API credits and on-demand GPUs
• Conference support: Costs may be covered if research is accepted by a reputable conference
• Extension funding: Possible for up to six months on a case-by-case basis
• Application deadline: September 6, 2026, at 11:59 PM Eastern Time
• Current status: Fall 2026 application deadline passed
• Official information: CBAI AI Safety Research Fellowship

About the Cambridge Boston Alignment Initiative

The Cambridge Boston Alignment Initiative, commonly known as CBAI, is a registered United States 501(c)(3) public charity supporting research and talent development related to AI safety, alignment, governance and associated risks.

The organisation uses the academic and professional ecosystem surrounding Cambridge and Boston to connect emerging researchers with experienced mentors, research groups, universities and AI safety organisations.

CBAI works with researchers and groups connected to institutions such as:

• Harvard University
• Massachusetts Institute of Technology
• Northeastern University
• Boston-area universities and research groups
• Institute for Responsible Superintelligence
• Goodfire
• AVERI
• METR
• Robocurve
• GovAI
• MIT AI Risk Initiative
• Institute for Law & AI

CBAI is not officially affiliated with Harvard University or MIT. However, it works with research groups and laboratories associated with these institutions and fiscally sponsors and operates certain student-led AI safety groups.

Purpose of the AI Safety Research Fellowship

The fellowship is intended to develop researchers who can contribute to the difficult technical, institutional, legal, political and economic questions created by advanced AI systems.

AI safety research examines how increasingly capable artificial intelligence systems can be developed, evaluated and governed without exposing society to unacceptable risks.

These risks can include:

• Systems behaving in ways that conflict with human intentions
• Failures that are difficult to identify or interpret
• Vulnerabilities in highly capable AI models
• Unsafe interaction among multiple autonomous systems
• Inadequate evaluation before deployment
• Misuse of advanced AI capabilities
• Institutional failure to manage AI development
• Concentration of technological and economic power
• Disruption of employment and economic systems
• Weak international coordination
• Catastrophic consequences from uncontrolled advanced systems

The programme gives fellows protected time, financial support, expert mentorship and computing resources to investigate these questions seriously.

Financial and Practical Support

US$15,000 Fellowship Stipend

Every selected participant will receive a stipend of US$15,000 for the ten-week programme.

CBAI states that fellows are not employed by the organisation. However, the stipend is characterised as taxable income. Participants are responsible for understanding their own tax and immigration obligations, as CBAI does not provide individual tax or legal advice.

Accommodation for Relocating Fellows

CBAI will arrange housing for accepted participants who need to relocate to the Cambridge-Boston area.

The current Fall 2026 benefits section states that accommodation may be provided through:

• Airbnb properties
• Short-term corporate rentals
• Other suitable temporary housing arrangements

The objective is to allow fellows to participate without having to independently secure expensive short-term accommodation in Cambridge.

Candidates who prefer to arrange their own housing should contact CBAI to determine whether reimbursement or an alternative arrangement may be available.

Free Weekday Meals

Participants will receive:

• Free weekday lunches
• Free weekday dinners
• Snacks
• Coffee
• Other beverages

The meal provision supports the programme’s intensive structure and encourages interaction among fellows, mentors, researchers and visiting speakers.

Dedicated Research Workspace

Fellows will receive twenty-four-hour access to a dedicated office in Harvard Square, located close to Harvard Yard.

The workspace is intended to give participants a professional environment for:

• Focused research
• Collaborative discussions
• Meetings with research managers
• Mentor sessions
• Paper-reading groups
• Research workshops
• Informal peer learning

Substantial Compute and Research Support

Each participant is expected to receive approximately US$1,500 per week in computing support.

Over ten weeks, this represents a nominal value of approximately US$15,000 per fellow, although it is not a cash payment. Support will be provided through resources such as:

• API credits
• Access to on-demand GPUs
• Computing infrastructure required for empirical research
• Additional resources for projects with unusually intensive computing requirements

CBAI states that the amount is flexible for empirical projects requiring heavier computational resources.

This support can be particularly valuable for research involving model evaluations, interpretability experiments, multi-agent systems, robustness testing or other computationally intensive work.

Conference Support

If a fellow’s paper is accepted by a reputable academic conference, CBAI may cover costs associated with participation.

Relevant conferences and workshops may include:

• Neural Information Processing Systems
• International Conference on Machine Learning
• International Conference on Learning Representations
• Specialist AI safety workshops
• Relevant policy and governance conferences

Conference support is conditional upon acceptance and should not be treated as an automatic cash benefit.

Extension Funding

Exceptional projects may continue beyond the initial ten-week programme.

CBAI may provide extension funding for up to six additional months on a case-by-case basis. The organisation reports that it has historically supported more than 75% of extension requests.

However, an extension is not guaranteed. Decisions are likely to depend on factors such as:

• Research progress
• Quality and importance of the work
• Mentor support
• Availability of funding
• Potential for publication or policy impact
• A credible plan for continued research

Research Track One: Technical AI Safety

The Technical AI Safety track focuses on reducing severe and potentially catastrophic risks from advanced AI systems.

Possible research areas include:

• AI alignment strategies
• Mechanistic interpretability
• Model transparency
• Robustness and reliability
• Formal verification
• Multi-agent safety
• Scalable oversight
• Model evaluations
• Monitoring advanced systems
• Detection of deceptive or undesirable behaviour
• Methods for keeping advanced systems under meaningful human control

This track may involve experimental work, theoretical research, programming, model analysis or development of new evaluation methods.

Research Track Two: AI Governance

The AI Governance track examines how public institutions, companies and international organisations should regulate and manage advanced AI.

Research may address:

• Domestic AI policy frameworks
• Regulatory approaches
• International coordination
• Institutional design
• Governance of frontier AI developers
• Risk-management requirements
• Regulatory oversight
• Accountability mechanisms
• International agreements
• Responsible deployment policies
• Public-sector AI capacity
• Governance responses to catastrophic or systemic risk

This track may be particularly relevant to researchers working in public policy, political science, international relations, law, public administration or technology governance.

Research Track Three: Technical Governance

Technical Governance connects technical understanding with public policy and institutional oversight.

Potential research areas include:

• Compute governance
• Monitoring access to advanced computing resources
• Model evaluations
• Technical standards
• Verification mechanisms
• Reporting and auditing requirements
• Hardware-enabled compliance systems
• Institutional mechanisms for AI safety
• Technical requirements for international agreements
• Governance of training and deployment thresholds

This track is valuable because effective AI regulation often requires mechanisms that can be measured, monitored and technically verified.

Research Track Four: Economics of Transformative AI

The Economics of Transformative AI track investigates the economic causes and consequences of advanced artificial intelligence.

Relevant subjects include:

• Economic growth driven by AI
• Productivity effects
• Labour-market transformation
• Employment displacement
• Changes in wages and skill demand
• Industrial strategy
• Market concentration
• Competition among AI developers
• Distribution of AI-generated economic benefits
• Political economy of AI governance
• Economic incentives influencing safety decisions
• Government responses to rapid technological change

Projects may use empirical analysis, economic modelling, policy analysis or combinations of these methods.

Individual Mentorship

Fellows will normally receive approximately one to two hours of individual mentorship each week.

Mentors are external specialists matched to the fellow’s particular research area. Their role includes:

• Providing specialist subject knowledge
• Helping refine the research question
• Advising on methods
• Reviewing preliminary findings
• Challenging assumptions
• Providing substantive feedback
• Supporting decisions about research direction
• Helping improve the final output

Mentor matching is a major component of the selection process. In rare cases, CBAI may issue a conditional offer to a strong applicant who has not yet been selected by a listed mentor. The offer would remain conditional until an appropriate listed or externally approved mentor agrees to supervise the candidate.

Role of the Research Manager

Each fellow will also work with an in-house research manager.

The research manager’s role differs from that of the external mentor. Research managers generally focus on:

• Project planning
• Progress monitoring
• Setting milestones
• Maintaining research momentum
• Identifying operational obstacles
• Clarifying research direction
• Improving research habits
• Supporting professional development
• Connecting the project with relevant CBAI resources

Depending on the research manager’s expertise, they may also provide direct research advice.

Fall 2026 Mentors

The Fall 2026 mentor pool listed by CBAI includes researchers, professors, technical specialists, policy professionals and organisational leaders.

Mentor	Listed affiliation or position
David Bau	Assistant Professor, Northeastern University
Dan Braun	Member of Technical Staff, Goodfire AI
Stephen Casper	Assistant Professor, Harvard Kennedy School
Michael Chen	AI Science Advisor, California Council on Science and Technology
Jay Chooi	CEO and Co-founder, Robocurve
Oliver Clive-Griffin	Member of Technical Staff, Goodfire AI
Shi Feng	Principal Investigator, Praxis; Assistant Professor, George Washington University
Zach Furman	Research Lead, Iliad
Samuel Gunn	Researcher, Institute for Responsible Superintelligence
Nikola Jurkovic	Member of Technical Staff, METR
Adam Tauman Kalai	CSO, Institute for Responsible Superintelligence
Benno Krojer	Postdoctoral Researcher, Northeastern University
Sean McGregor	Co-founder, AVERI; Founder, AI Incident Database
Dylan Hadfield-Menell	Associate Professor, MIT
James Mickens	Gordon McKay Professor of Computer Science, Harvard University
Michael Noetel	Senior Researcher, MIT AI Risk Initiative
Hadas Orgad	Research Fellow, Harvard University
Patricia Paskov	Director of Standards, AVERI
Paul Riechers	Co-founder, Simplex; Research Lead, Astera Institute
Alexander Saeri	Project Director, MIT AI Risk Initiative
Peter Salib	Associate Professor of Law, University of Houston; Founder and Executive Co-director, Center for Law and AI Risk
Adam Shai	Co-founder, Simplex; Co-lead, Astera
Jonathan Shafer	Postdoctoral Associate, MIT; Researcher, RESI; Incoming Assistant Professor, Weizmann Institute of Science
Peter Slattery	Research Scientist, MIT AI Risk Initiative
Hidenori Tanaka	Group Leader, Physics of Intelligence Group at Harvard University
Charles Teague	CEO, Meridian Labs
Kevin Wei	Research Scholar, GovAI
Gabriel Weil	Assistant Professor, University of Houston; Non-Resident Senior Fellow, Institute for Law and AI
Jonathan Zittrain	Professor of International Law, Harvard Law School

Mentor availability and matching may depend on each candidate’s research interests, skills and the mentor’s capacity.

Speaker Events and Professional Community

The programme includes a weekly speaker series featuring researchers working in AI safety, governance and related fields.

Previously featured speakers include:

• Joe Carlsmith of Anthropic
• Ekdeep Singh Lubana of Goodfire
• Stewy Slocum of xAI
• Max Nadeau of Coefficient Giving

Additional programme activities may include:

• Research workshops
• Policy and technical discussions
• Networking events
• Wargames
• Paper-reading groups
• Informal dinners
• City outings
• Board-game evenings
• Events with Cambridge-area research communities
• Opportunities to meet visiting researchers

Fellows whose mentors work in local laboratories may be expected to use those laboratory facilities approximately once or twice per week, subject to the relevant arrangement.

A Typical Fellowship Day

A normal day is expected to begin at the Harvard Square workspace, where fellows concentrate on their individual research projects.

Around lunchtime, participants may attend a programme speaker session involving an in-person or virtual expert. Afternoons generally provide longer periods for focused research, experimentation, analysis and writing.

Evenings may include group dinners followed by activities such as:

• Paper discussions
• Collaborative research sessions
• Informal networking
• Board games
• Community events

Throughout the week, fellows may also participate in:

• Individual check-ins with their research managers
• Meetings with CBAI team members
• Coffee conversations with other fellows
• Mentor meetings
• Research workshops
• Events involving nearby university groups

The programme therefore combines independent research with structured mentorship and an active professional community.

Expected Research Outputs

Fellows are expected to produce work suitable for public sharing.

Possible outputs include:

• Research reports
• Blog posts
• Interactive demonstrations
• Preprints
• Academic papers
• Policy reports
• Submissions to ICLR, ICML or NeurIPS
• Workshop submissions
• Technical or governance research tools

CBAI does not impose one rigid format on every project. The appropriate output will depend on the research question, methods and track.

The organisation intends to support participants in identifying suitable conferences, workshops and publication opportunities.

Previous Fellowship Outcomes

Members of CBAI’s inaugural fellowship cohort have reportedly:

• Joined organisations including Goodfire and Redwood
• Established an independent research group
• Had work accepted at NeurIPS and ICLR
• Shared research reports with policymakers in Washington, D.C.
• Continued collaborating with other researchers
• Developed ongoing relationships with mentors

These examples demonstrate that the programme is intended to support concrete research and career outcomes rather than functioning only as a short introductory course.

Relationship With Harvard and MIT

CBAI is not formally affiliated with Harvard University or MIT.

It nevertheless works closely with research groups and laboratories connected with Harvard, MIT, Northeastern and other Boston-area institutions.

CBAI also fiscally sponsors and operates student groups associated with AI safety activities at Harvard and MIT. Depending on their projects and mentors, some fellows may receive affiliate access to relevant university facilities.

Students enrolled at institutions participating in BorrowDirect may separately qualify for library cards offering access to Harvard and MIT libraries. This depends on the student’s university arrangements and is not an automatic benefit for every fellow.

Application and Selection Process

The selection process consists of four principal stages.

Stage One: General Application

Candidates first complete the online application form covering their background, experience, research interests, motivation and future career plans.

Stage Two: CBAI Interview

Shortlisted applicants are invited to a 15-minute interview with the CBAI team.

Stage Three: Mentor-Specific Assessment

Depending on the proposed mentor and research track, applicants may be asked to complete:

• A mentor-specific written question
• A short research task
• A technical test
• A coding screen
• Another assessment relevant to the proposed project

Not every applicant will necessarily receive the same assessment.

Stage Four: Mentor Interview

Candidates who progress will interview with a prospective mentor. Mentor agreement is important because the programme is structured around individual research supervision.

What Makes a Competitive Application

According to CBAI, successful applicants generally demonstrate:

• Clear alignment with the programme’s goals
• A well-developed interest in AI safety
• Strong motivation to conduct research
• A credible career plan for the period after the fellowship
• Relevant academic, professional or independent experience
• Skills connected to the proposed research
• A clear explanation of why this programme is the right next step
• An understanding of the impact they hope to achieve
• A realistic plan for using the fellowship to advance their career

Applicants should be specific. Generic claims about being interested in artificial intelligence are unlikely to distinguish an application.

A stronger application explains:

1. Which AI safety problem the applicant wants to address.
2. Why that problem matters.
3. What relevant work the applicant has already completed.
4. Which skills the applicant can contribute.
5. What skills the applicant needs to develop.
6. How the fellowship connects with a serious post-programme career plan.
7. What research or policy impact the applicant hopes to achieve.

Attendance and Time Commitment

The default expectation is full-time, in-person attendance for the entire ten-week programme.

CBAI does not ordinarily permit part-time participation.

Virtual attendance is also not generally available. Exceptions may be considered in unusual circumstances, including situations where a disability makes physical attendance extremely difficult.

Requests for partial or remote attendance are assessed individually and should not be assumed to receive approval.

Referral Incentive

CBAI states that a person who refers a candidate who is ultimately selected may receive a US$500 Amazon gift card.

The referral incentive is separate from the fellow’s stipend and does not affect the applicant’s stated selection requirements.

Application Deadline and Current Status

The application deadline was:

September 6, 2026, at 11:59 PM Eastern Time

The same deadline was published through CBAI’s official announcement.

As this deadline has passed and no official extension was identified, the Fall 2026 programme should not be promoted as currently accepting applications.

Upcoming Spring Cycle and J-1 Visa Sponsorship

CBAI states that details of its next Spring cycle are expected to be released in late September.

The organisation also states that it expects to be able to sponsor J-1 visas beginning with the Spring cycle. This is an important change because the Fall 2026 programme does not offer visa sponsorship.

Future applicants should not assume that the Fall 2026 dates, stipend, benefits or application requirements will automatically remain identical for the Spring programme. All new details should be verified when the next call is published.

Interested candidates can complete the CBAI upcoming-cohort update form.

Important Information to Confirm

The official page contains some FAQ language referring to an earlier summer cycle. Applicants considering a future cohort should therefore verify:

• The final type and location of accommodation
• The number of available places
• Whether the AI Safety and AIxBiosecurity programmes will run together
• The precise Spring programme dates
• The Spring stipend and computing allowance
• J-1 visa eligibility and sponsorship procedures
• Whether any virtual or accessibility-based arrangements are available

Questions may be sent to fellowships@cbai.ai.

Conclusion

The CBAI Fall AI Safety Research Fellowship 2026 offers an unusually comprehensive package for emerging researchers seeking to build serious careers in technical AI safety, AI governance, technical governance or the economics of transformative AI.

Its combination of a US$15,000 stipend, provided accommodation, free weekday meals, approximately US$1,500 per week in computing resources, individual mentorship, research-management support, conference assistance and access to the Cambridge-Boston AI research community makes it substantially more than a conventional academic training programme.

Participants are expected to conduct meaningful research, produce publicly shareable outputs and use the experience to advance a credible long-term career in AI safety. The programme is intensive, full-time and primarily in person, with strict United States work-authorisation requirements for the Fall 2026 cycle.

Applications for the Fall programme closed on September 6, 2026. Candidates without current United States work authorisation should monitor the forthcoming Spring call, for which CBAI expects to introduce J-1 visa sponsorship.

For more global opportunities, click HERE.