Let me be honest with you. Most “research fellowships” are glorified internships where you spend your summer making slides for someone else’s presentation. The CBAI Summer Research Fellowship in AI Safety 2026 is not that.
This is nine weeks in Cambridge, Massachusetts — Harvard Square, to be precise — working on one of the most consequential problems of our time: making sure advanced AI systems don’t go sideways. And they’re paying you $10,000 to do it. Plus housing. Plus meals. It’s the kind of opportunity that sounds too good to be true, but isn’t.
The deadline is April 12, 2026. So if you’re reading this and procrastinating, stop.
What Exactly Is the CBAI Fellowship?
The Cambridge Boston Alignment Initiative (CBAI) runs this flagship program every summer. Think of it less like a university course and more like being dropped into a small, high-functioning research lab — except your colleagues have affiliations with MIT, Harvard, RAND, and Goodfire.
The program runs June 8 – August 10, 2026 (nine weeks) and is designed for students and researchers who are serious about AI safety. Not curious-about-it serious. Actually working on it serious.
Fellows in the inaugural cohort have gone on to join Goodfire and Redwood Research, presented findings to policymakers in Washington D.C., and had work accepted at NeurIPS and ICLR. That’s not a bad nine weeks.
What Does the Fellowship Actually Cover?
Here’s the full picture — no fine print:
| Benefit | Details |
|---|---|
| Stipend | $10,000 for the full nine weeks |
| Housing | Free (Harvard dorms or equivalent) |
| Meals | Free weekday lunch and dinner |
| Compute Support | Up to $10,000 in API credits and GPU access |
| Workspace | 24/7 access to office space in Harvard Square |
| Conference Support | Funding for submitting to NeurIPS, ICLR, etc. |
| Extension Fellowship | Up to 6 months of additional funding (if eligible) |
Yes, that’s potentially $20,000 in total value when you factor in compute support. For a nine-week summer program, that’s remarkable.
The Research Tracks — Choose Your Focus
CBAI isn’t a one-size-fits-all program. You apply to one of four research tracks depending on your background and interests:
1. Technical AI Safety Interpretability, alignment strategies, scalable oversight, robustness. This is the track for people who want to get into the technical weeds of how models behave — and misbehave.
2. AI Governance Policy frameworks, international coordination, regulatory design. If you think the real lever for AI safety is in institutions and law rather than code, this is your lane.
3. Technical Governance The bridge between the two above — model evaluation, compute governance, safety standards. Arguably the most interdisciplinary and in-demand specialization right now.
4. Economics of Transformative AI Labor markets, productivity impacts, the macroeconomics of AI deployment. Underexplored, critically important, and wide open for new research.
Who Should Apply (And Who Definitely Should)
The eligibility list is broader than you might expect:
- Undergraduate, Master’s, and PhD students
- Postdoctoral researchers
- Recent graduates pivoting into AI safety
- Anyone 18+ with US work authorization (OPT/CPT accepted; no visa sponsorship)
Here’s something worth noting: CBAI has explicitly said it believes the most promising undergraduates are severely neglected in the AI safety field. Some mentors have undergraduate-specific quotas. If you’re an undergrad who’s been told “come back when you have a PhD,” this program is actively making the opposite argument.
The Mentors Are the Real Story
You can talk about stipends and housing all you want, but what makes a research fellowship genuinely valuable is who you learn from. CBAI’s 2026 mentor list is exceptional:
- Max Tegmark — Professor of Physics, MIT (yes, that Tegmark)
- Dylan Hadfield-Menell — Associate Professor of EECS, MIT CSAIL
- Andreea Bobu — Assistant Professor, MIT CSAIL
- Ekdeep Singh Lubana — Member of Technical Staff, Goodfire
- Kevin Wei — Adjunct Staff, RAND
- Peter Slattery — Lead, MIT AI Risk Repository
Weekly one-on-one sessions with researchers at this level isn’t just professionally useful. It’s the kind of mentorship that reshapes how you think about problems.
Previous speaker series guests have included Joe Carlsmith from Anthropic and Stewy Slocum from xAI — so the network you’re plugging into here is genuinely frontier-level.
How to Apply — Step by Step
The application process has four stages, and CBAI reviews on a rolling basis (meaning earlier is better, not just before the deadline):
- General Application Form — Research interests, motivation, background
- 15-Minute Initial Interview — A quick fit-check with the CBAI team
- Mentor-Specific Screening — Could be a test task, code review, or written questions depending on your track
- Interview with Your Potential Mentor — The real conversation about your research direction
For the application itself, come with clarity. Know which track you’re applying to and why. Have a GitHub, portfolio, or writing samples ready. And if you’re applying to a technical track, make sure your code demonstrates actual depth — not just familiarity with the tools.
Pro tip from people who’ve been through similar programs: Mentors are looking for “research taste” — the ability to identify interesting, neglected questions — not just technical credentials. Show you’ve thought about what’s worth working on, not just what you’re capable of doing.
Frequently Asked Questions
Can international students apply? Yes, if you already have US work authorization. OPT and CPT are accepted. CBAI cannot sponsor visas.
What does a typical day look like? Research work, weekly workshops, speaker events, and one-on-ones with mentors. It runs like a research lab, not a classroom.
Is CBAI affiliated with Harvard or MIT? CBAI is an independent 501(c) nonprofit, but operates in close proximity to both institutions and draws heavily from their research communities.
Can I attend part-time or virtually? The program is designed to be in-person and full-time. Part-time or virtual participation is not the standard format.
What if no mentor selects me? CBAI has addressed this in their FAQ — reach out to them directly at emre@cbai.ai for specifics on how unmatched applicants are handled.
What’s the difference between a research manager and a mentor? Your mentor is an external researcher guiding your specific project. Your research manager is CBAI’s in-house support — think of them as your day-to-day resource for navigating the program.
The Bottom Line
If you’re working in AI safety — or trying to break into it — the CBAI Summer Research Fellowship is one of the most substantive opportunities available in 2026. The funding is real, the mentors are legitimate, and the research output matters.
The deadline is April 12, 2026.
Apply at your earliest convenience. Rolling review means waiting costs you. And if you refer someone who gets accepted, CBAI offers a $100 Amazon gift card — so share this with the right people.
For questions, contact: emre@cbai.ai
Apply here: cbai.ai