Berkman Klein Center Fellowship at Harvard: Up to $75,000 for a Year of Public-Interest AI Research

Berkman Klein Center Fellowship at Harvard: Up to $75,000 for a Year of Public-Interest AI Research

The Berkman Klein Center for Internet & Society at Harvard University runs one of the few residential fellowships that will fund a lawyer, a philosopher, a machine-learning engineer, and a civil-society organizer to work on the same set of questions in the same building. Its annual open call is the main door in, and it pays real money: fully funded fellows receive up to $6,250 per month, capped at $75,000 for a twelve-month appointment.

Status first, because it determines what you should do today. The most recent annual open call covered two appointment periods — January through August 2026, and September 1, 2026 through August 31, 2027 — and it closed on Friday, December 5, 2025 at 11:59 p.m. ET. The Center’s fellowship page now states plainly that the call has closed and that applications are no longer being accepted. As of this writing, BKC has not published an opening date or a deadline for the next annual cycle. This page is therefore a preparation guide and a monitoring reference, not an active application link.

That is still worth your attention. This is a recurring annual program with a stable structure, published selection criteria, and a document list that does not change much year to year. If the next round follows the pattern of the last one, you have a few months of runway — and the thing that separates a strong BKC application from a weak one is almost entirely built before the call opens.

Key details

What the fellowship actually funds

BKC fellowships are not open-ended sabbaticals. The call is explicit that a fellow is expected to produce at least one significant public output that impacts and/or informs the scholarly, public, and/or policy debates. The Center names several acceptable forms: technical prototypes, open-source tools, writing, and workshops. That breadth is the point. A well-argued law review article and a released evaluation harness are both legible deliverables here, which is unusual among research fellowships that quietly expect only papers.

The second expectation is community engagement. Fellows are asked to work with faculty, staff, students, and other members of the BKC and wider Harvard communities, and to partner closely with specific topical workstreams rather than orbiting the Center at a distance. The FAQ says fellows “work with specific topical workstreams and partner closely with other fellows, faculty, and staff.” Applicants who show excitement about working with and mentoring students are viewed favorably.

This shapes what a good project proposal looks like. A project that could equally well be done alone at your home institution is a weak fit. A project that gets materially better because you are next to a computer scientist, a platform-governance researcher, and a clinical law program is a strong one.

The five research priorities in the last call

The most recent open call named five thematic priorities. These are the clearest available signal of where the Center is putting its attention, and they are worth reading as a research agenda rather than a list of keywords to sprinkle into a proposal:

1. Agentic AI — governance of autonomous AI agents that operate across systems without human mediation at each step.
2. Language model interpretability — understanding how and why language models function.
3. Benchmarking AI systems beyond measures of intelligence — evaluating dimensions such as agency and metacognition rather than capability alone.
4. AI and the human experience — effects on relationships, cognition, spirituality, and social-emotional development.
5. Bridging the AI triad — building dialogue among accelerationists, safetyists, and skeptics.

Two things stand out. First, three of the five are technical or partly technical, which is why the Center now runs an explicit technical track (below). Second, priorities four and five are unapologetically humanistic and deliberative — they are asking for people who can write and convene, not only measure. If your work sits at either pole, do not assume the other pole is what they really want.

Note that these were the priorities for the closed round. Treat them as a strong prior for the next call, not a guarantee; confirm against the new call text when it is posted.

Who fits, and who does not

The open call is aimed at people who already have a research practice. Specifically:

• Faculty, anywhere from assistant through full professor.
• Post-doctoral scholars who have recently completed a terminal degree.
• Practitioners from industry, government, and civil society — this is a genuine pathway, not a courtesy mention.
• International applicants, with visa sponsorship available.

Interdisciplinary backgrounds are explicitly encouraged. Concurrent fellowships elsewhere at Harvard are permitted where there is enough flexibility to meet both commitments.

Students are not eligible through the annual open call. If you are mid-PhD, this is not your route; look instead at BKC’s student programs, internships, and discussion groups, which are administered separately. Similarly, faculty associate and affiliate roles are appointed by invitation only — there is no independent application form for them, although the Center may consider an open-call applicant for one of those designations based on the same materials.

The technical research and engineering track

For technical roles, the call sets out a sharper profile: ideally a Ph.D. in computer science or a related technical field, with demonstrated expertise in Python and modern AI/ML frameworks such as PyTorch or JAX, plus peer-reviewed publications or substantial technical contributions. Research engineer candidates are asked to meet at least three of six listed qualifications, covering areas including infrastructure and agent-framework experience.

If you are applying on this track, your work samples should be code and artifacts, not only PDFs. A maintained repository with real users, an evaluation suite others have adopted, or a prototype that demonstrably works will do more than a fourth workshop paper.

Residency, remote work, and the honest calculation

BKC strongly encourages fellows to be in residence in Cambridge, Massachusetts for the majority of their appointment. Non-residential arrangements are considered individually, and the FAQ is candid that remote fellows are possible but discouraged; the Center asks applicants to detail their residence plans in the application itself.

Do that arithmetic before you write anything. A $75,000 twelve-month stipend is a serious sum, but Cambridge is an expensive place to live, and the stipend is the funding mechanism, not a salary with a benefits package attached. Fellows on leave from a salaried post, fellows combining BKC support with sabbatical pay or an external grant, and fellows early in their careers with modest fixed costs are all in reasonable positions. A mid-career practitioner relocating a household on the stipend alone should model it carefully rather than optimistically.

The Center also notes that offers are extended based on a number of factors including funding requirements, and asks applicants to communicate their funding needs clearly. This is an invitation to be specific rather than modest. Stating that you have partial external support, or that you need the full amount to make the year possible, is useful information for the committee — not a mark against you.

Required materials

The last call asked for the following. All uploads must be PDFs, with individual files capped at 5 MB.

• CV.
• Cover letter, 1–2 pages.
• Project proposal, 2–3 pages.
• Work samples, one to three, combined into a single PDF.
• Two professional references, with contact details.

The proposal is the load-bearing document. Two to three pages is short enough that a vague framing has nowhere to hide. A workable structure: the question and why it is unresolved; what you will actually make during the appointment; what makes BKC the right place rather than a nice place; what the public output will be and who it is for; and what you will do in the first sixty days. Reviewers can tell the difference between a plan and an aspiration.

The cover letter should carry the parts of your story the CV cannot — why now, what changed in your work, what you want to be different about your practice afterward.

How selection works

Applications are reviewed by senior staff and faculty directors, who may consult other members of the community. The Center says it seeks candidates with novel and unique approaches, and lists the qualities it weighs: ambition, relevant achievements, collegiality, curiosity, intellectual rigor, adaptability, and commitment to the public interest.

Collegiality and adaptability appearing on a selection list is not decoration. This is a cohort program where fellows are placed alongside people trained very differently from them; someone who cannot productively disagree with an economist or take a technologist’s objection seriously will have a poor year and drag on everyone else’s. The Center’s stated criteria for technical fellows make the same point from the other direction — independent research leadership aligned with BKC’s AI agenda, interdisciplinary communication skills, student mentoring, and dedication to public-interest, open-source research.

Cohort size is not fixed. The FAQ says the Center determines it each year based on applicant characteristics, its capacity to support fellows, and the Center’s needs. Do not try to reverse-engineer an acceptance rate; there is not a stable one to find.

Common mistakes

• Writing a proposal that ignores the named priorities. You do not have to sit dead-center in one of the five themes, but you should be able to say in one sentence how your work touches BKC’s current agenda.
• Treating the public-output requirement as boilerplate. Name the artifact. “A policy brief and an accompanying open dataset released by month nine” beats “I will disseminate my findings.”
• Being vague about residency. The application asks about it directly. An unaddressed plan to be mostly elsewhere reads worse than a clearly reasoned one.
• Applying as a student. The open call does not cover students, and an ineligible application is simply lost effort.
• Ignoring the file constraints. PDFs only, 5 MB per file. Every cycle, strong candidates scramble at the deadline because a work-sample PDF is oversized.
• Waiting for the call to ask for references. Two professional references with contact details are required at submission. Ask in advance.

What to do between now and the next call

The annual open call is not the only way in. The Center makes periodic calls for applications for project-specific fellowships throughout the year, tied to particular initiatives — the Institute for Rebooting Social Media and its Assembly Fellowship, the Applied Social Media Lab, and the Visiting Scholars Program among them. These are announced on an irregular schedule, which makes monitoring worthwhile rather than optional.

A concrete plan:

1. Subscribe to The Buzz, BKC’s newsletter, which is where the Center says new fellowship opportunities are announced. Check the fellowship page directly every few weeks in the autumn.
2. Draft the proposal now. Two to three pages of genuinely sharp writing takes longer than people expect, and the version written in the last week before a December deadline is rarely the good one.
3. Assemble the work-sample PDF now. Pick the samples that show range and rigor, not just recency. Combine and compress them once, properly.
4. Line up references early, and give them the proposal so their letters describe the same project you do.
5. Read current fellows’ output. The Center publishes its fellows’ work; a few hours reading it tells you more about what lands there than any application guide, including this one.

Frequently asked questions

Is the fellowship open right now? No. The 2026 and 2026–2027 open call closed on December 5, 2025, and the fellowship page confirms applications are no longer accepted. No date for the next annual call has been published.

When will the next call open? Not confirmed. The last cycle ran to a December deadline, so a late-autumn call is a reasonable planning assumption — but BKC has not stated this, and you should verify against the official page rather than plan around an inference.

Is the stipend guaranteed at $75,000? No. That is the ceiling for a twelve-month appointment, at up to $6,250 per month. Amounts are determined case by case, and some appointments are externally funded.

Can I hold another position at the same time? Concurrent fellowships elsewhere at Harvard are permitted if there is sufficient flexibility to meet both sets of obligations. For outside commitments, raise it explicitly in your application rather than after an offer.

Do I need to move to Cambridge? Residency for the majority of the appointment is strongly encouraged. Non-residential fellowships are considered individually, and you are asked to explain your plan in the application.

Can international applicants apply? Yes, and visa sponsorship is available.

Are there other ways to become affiliated with BKC? Faculty associate and affiliate roles are appointed by invitation; there is no separate application. Project-specific fellowship calls are posted periodically and are a genuine alternative route.

Official links and contact

• Fellowship program overview: https://cyber.harvard.edu/getinvolved/fellowships
• Fellowship program FAQ: https://cyber.harvard.edu/getinvolved/fellowships/faq
• Most recent open call (closed, useful as a reference for materials and priorities): https://cyber.harvard.edu/page/open-call-fellowship-applications-2026-and-2026-2027
• Berkman Klein Center for Internet & Society, 1557 Massachusetts Avenue, 5th Floor, Cambridge, MA 02138
• Phone: (617) 495-7547 — Email: [email protected]

Verify every date, amount, and requirement against the new call text when it is published. The structure of this program has been stable, but the thematic priorities and the stipend ceiling are set per cycle, and the official page is the only authority on both.