PIBBSS Fellowship 2026–2027: A Fully Funded 3-Month AI Safety Research Program in Cape Town With a $3,000 Monthly Stipend

PIBBSS Fellowship 2026–2027: A Fully Funded 3-Month AI Safety Research Program in Cape Town With a $3,000 Monthly Stipend

Most funding for artificial intelligence safety flows to computer scientists and machine learning engineers. The PIBBSS Fellowship takes a deliberately different route. Run by Principles of Intelligence (the organization formerly known as PIBBSS, short for Principles of Intelligent Behavior in Biological and Social Systems), the fellowship invites researchers from mathematics, the natural sciences, the social sciences, and the humanities to bring their own disciplinary tools to the problem of making advanced AI systems safe. For the inaugural 2026–2027 winter cohort, the program offers a fully funded, roughly three-month research placement in Cape Town, South Africa, with a monthly stipend of $3,000 and mentorship from experienced AI safety researchers.

This guide is built from the official program page at princint.ai rather than a reposted announcement. It explains what the fellowship pays, who it is designed for, how the four-stage selection works, and how to put together an application before the July 20, 2026 deadline. If you are a researcher who has always suspected your field has something to say about intelligent systems and their risks, but you have never had a structured, funded way to test that hunch, this is one of the clearest on-ramps available.

Key Details at a Glance

Treat this table as a planning map. The most important number is the deadline: the written application is due on July 20, 2026, and the interview rounds that follow it are compressed into the weeks before the fellowship begins in November.

What the Fellowship Offers

The core of the fellowship is time and support to do serious research. For roughly three months, fellows work full time on a project at the intersection of their own discipline and AI safety, under the mentorship of an experienced researcher matched to their interests. Around that core, PIBBSS wraps a structured program: pre-program reading groups to get everyone oriented, multi-day opening and closing retreats, and a final symposium in which fellows present their work.

The financial package is designed so that money is not the reason you say no. Fellows receive a $3,000 monthly stipend for the duration of the program. On top of the stipend, the program provides a private bedroom in shared accommodation, meals on workdays at the office, and one return flight to Cape Town. For international fellows, PIBBSS also issues visa support letters, which can make the difference between a smooth entry and a bureaucratic dead end. In practice, this means an accepted fellow can relocate to Cape Town, live and eat comfortably, and focus entirely on the research rather than on covering costs.

There is also a longer-term payoff that does not show up on the stipend line. PIBBSS reports that its 73 alumni have gone on to organizations including Anthropic, Google, Harvard, and Oxford, as well as dedicated AI safety labs. The fellowship functions as a credible entry point and a network: mentors, co-fellows, and the wider AI safety community you meet during the program often shape where you go next. For researchers trying to pivot into a field that can be hard to break into from the outside, that access is a substantial part of the value.

Who Should Apply

The fellowship is built for people with real research experience who want to redirect it toward AI safety. PhD students and postdoctoral researchers are the central audience, but the program is explicit that it considers applicants of all credentials, and what it is really screening for is substantial research experience and genuine motivation rather than a particular degree. If you have spent years learning to formulate questions, gather evidence, and defend conclusions in some rigorous field, you have the raw skills the program wants — the fellowship exists to point those skills at a new target.

The list of welcomed disciplines is unusually broad on purpose: mathematics, neuroscience, cognitive science, dynamical systems, physics, philosophy, political and economic theory, ecology, evolutionary biology, linguistics, media studies, and the humanities all appear. The unifying idea is that intelligence, agency, coordination, and risk are studied across many fields, and that AI safety benefits from importing frameworks that machine learning alone does not supply. A philosopher who thinks carefully about agency, an ecologist who models complex adaptive systems, or a political theorist who studies how institutions constrain powerful actors can each bring something the field needs.

If you are already a full-time machine learning safety researcher with an established position, this particular program is probably not aimed at you — it is a transition and entry program. And if you are curious about AI but have little independent research experience, you may be better served by shorter introductory courses first, then applying to a future cohort once you have a body of work to point to. Reading the eligibility framing honestly before you invest in an application saves everyone time.

Eligibility and Fit in Detail

The formal bar is deliberately loose on credentials and firm on substance:

• Substantial research experience. You should be able to show that you can carry a research question from formulation to result. This is usually evidenced by a PhD, postdoctoral work, or an equivalent track record of independent research output.
• Motivation to contribute to AI safety. The program is not a general research residency. Reviewers want to see that you have thought about why advanced AI poses risks and how your field could help address them.
• A disciplinary lens. Your background in one of the welcomed fields is an asset, not a limitation. The strongest applications explain concretely how a specific concept or method from your field maps onto an AI safety question.
• Open to all nationalities. Applicants from every country are welcome, and the program provides flight support and visa letters precisely so that geography is not a barrier.

Fit matters as much as raw eligibility here. Because the program pairs each fellow with a mentor, part of what selectors are assessing is whether there is a plausible research direction and a mentor who could support it. That makes specificity in your application unusually valuable: a clearly articulated question that connects your expertise to a real AI safety problem is far more compelling than a general statement of interest.

The Application Process and Timeline

PIBBSS runs a four-stage selection. The first stage is a written application: a CV, a personal statement of 600 to 800 words, and optional work samples such as papers, code, or writing that demonstrate your research ability. Stages two through four are interview rounds, in which the team discusses your research interests, probes your thinking, and works to match you with an appropriate mentor. The interviews are as much about finding the right research pairing as they are about screening — a good match between fellow and mentor is central to how the program is designed.

The calendar is tight, so plan backward from the deadline:

• Now through mid-July 2026: decide on a research direction, draft and revise your personal statement, and assemble a current CV and any work samples.
• July 20, 2026 (23:59 Anywhere on Earth): the written application is due. Submit with time to spare rather than at the final minute.
• Weeks after the deadline: shortlisted applicants move through the interview rounds and mentor matching.
• November 2026 – February 2027: the fellowship runs, roughly three months, with reading groups beforehand and a closing symposium in spring 2027.

Because the fellowship begins in November, the months between the July deadline and the program start are compressed. If you are selected, you will likely need to arrange travel, visas, and any leave from your current work quickly, so it is worth thinking through the logistics before you apply rather than after.

Preparing a Strong Application

The personal statement is where most applications are won or lost, and 600 to 800 words is short enough that every sentence has to earn its place. A few principles help:

• Lead with a concrete question, not a general interest. “I want to work on AI safety” tells reviewers nothing. “I study how ecological systems remain stable under perturbation, and I want to test whether those stability concepts can formalize what we mean by a ‘corrigible’ AI system” tells them exactly what you would do and why you specifically could do it.
• Make the bridge from your field explicit. The program is betting that your discipline has transferable tools. Name the tool, name the AI safety problem, and show the connection.
• Show research maturity. Reviewers are assessing whether you can run an independent project. Reference work you have actually done — a thesis, a paper, a model you built — as evidence, and choose work samples that display rigor rather than volume.
• Be honest about what you do not yet know. You are not expected to already be an AI safety expert. Demonstrating that you understand the shape of the field and have a credible plan to learn fast is more persuasive than pretending to fluency you lack.

For work samples, quality beats quantity every time. One strong paper or a well-documented project that shows how you think will do more than a long list of minor outputs. If your best work is co-authored, be clear about your specific contribution.

Common Mistakes to Avoid

• Treating it as a generic research grant. The program is specifically about AI safety. Applications that never engage with why AI risk matters, or how the applicant would address it, read as off-target.
• Being vague about the research direction. Because the program matches fellows to mentors, a fuzzy proposal is hard to place. Specificity is a feature, not a risk.
• Overselling AI expertise. Claiming deep familiarity you do not have is easy to detect in interviews. Lean on your genuine strengths and show you can learn the rest.
• Ignoring the logistics. The fellowship is in-person in Cape Town from November. If travel, visas, or existing commitments would make that impossible, address how you would handle it rather than leaving it as an open question.
• Submitting at the deadline. Portals and connections fail at the worst moments. Finish a few days early.

Frequently Asked Questions

Is the fellowship fully funded? Yes. Fellows receive a $3,000 monthly stipend, private accommodation, workday meals, and one return flight to Cape Town, along with visa support letters.

Do I need a background in computer science or machine learning? No. The program specifically welcomes researchers from mathematics, the natural and social sciences, and the humanities. What matters is substantial research experience and a serious interest in AI safety.

Can I participate remotely? The fellowship is designed to be in-person in Cape Town, but remote participation is available for exceptional candidates. Assume in-person unless you have a strong case otherwise.

How long is the program and when does it run? It runs approximately three months, from November 2026 to February 2027, with pre-program reading groups and a closing symposium in spring 2027.

How many fellows are selected? Approximately 20 for the 2026–2027 winter cohort.

What is the deadline? The written application is due July 20, 2026, at 23:59 Anywhere on Earth.

Final Checklist and Official Links

Before you submit, confirm that you meet the experience bar, that your personal statement stays within 600 to 800 words and centers on a concrete research direction, that your CV is current, and that any work samples show your best rigorous work. Make sure you can realistically be in Cape Town from November 2026 to February 2027, or that you have a strong case for remote participation. Then submit ahead of the July 20, 2026 deadline rather than at it.

Always verify the latest dates, stipend, and requirements on the official program page before applying, since details can change between cohorts. The authoritative source is the Principles of Intelligence (PIBBSS) fellowship page: https://princint.ai/programs/fellowship/. Treat any third-party summary, including this one, as a planning aid and confirm the specifics directly with PIBBSS.