PIBBSS Fellowship 2026–2027: Historical Reference for the Closed Application Cycle in Cape Town

PIBBSS Fellowship 2026–2027: Historical Reference for the Closed Application Cycle in Cape Town

The PIBBSS Fellowship is an interdisciplinary research program run by Principles of Intelligence (PrincInt). It connects researchers from fields that study intelligence, complex systems, institutions, and living systems with AI safety research and mentorship. For the 2026–2027 Winter Fellowship, applications are closed. The fellowship itself is scheduled to run for about three months in Cape Town, from November 2026 to February 2027, so this is an archive of a closed recruitment round rather than an invitation to apply.

The official page gives July 20, 2026 as the application deadline and links to an expression-of-interest form for future opportunities. That form allows people to register interest in a future fellowship, affiliate work, events, or collaboration. It says PrincInt is preliminarily planning a fellowship every summer, but does not announce an application window, confirmed dates, or terms for a next cohort. The date below therefore records the real deadline for this closed round. Anyone interested in a later opportunity should use the form or watch the official program page for a new announcement.

Key facts from the 2026–2027 Winter Fellowship

These terms describe the published winter cycle. The official page has a clear closure notice near the top and a timeline that records July 20, 2026 as the deadline. Lower down, its application-process section still says applications are open and an application-form link remains on the page. Those fragments conflict with the closure notice and dated timeline; the official page does not say that late submissions are accepted or that applications remain open until the cohort fills. Treat this round as closed rather than using the leftover form link to infer that applications are still being accepted.

The official page also lists its “2026 Fellows.” Read together with the closure notice and the dated timeline, that section supports treating the winter recruitment as complete even though the program itself is scheduled for later in 2026 and early 2027. The page does not announce a replacement application cycle.

Program structure and research tracks

The fellowship was designed as full-time research at the intersection of each fellow’s expertise and AI safety. A remote pre-program reading group introduces AI risk topics, followed by opening and closing retreats with the cohort and researchers. During the research period fellows use shared office space, attend talks and community events, and work with a dedicated mentor. The program page describes a final symposium presentation in Spring 2027. These features make the program a mentored research placement with a cohort, rather than a general scholarship or an online course.

The published research agendas are varied. The Gradual Disempowerment track examines how human civilization might change if economic, cultural, or state systems no longer depend on human participation, and how people could retain meaningful agency. The page names post-AGI economics, culture, and governance as possible areas, and particularly welcomes economists, cultural-evolution and evolutionary-biology theorists, and researchers comfortable crossing disciplines.

The ASI Safety via AIXI track focuses on the incentives of safe and unsafe advanced AI architectures through algorithmic information theory, including agent foundations and implementation. Example topics include containment protocols, robust assistance games, generalization, and links between machine learning and algorithmic information theory. The Safe Pareto Improvement track, supported by the Center on Long-Term Risk, concerns ways of changing bargaining strategies so all parties can benefit. The page describes both conceptual questions about when agents prefer such improvements and empirical or evaluation work on model behavior and research automation; interested applicants were asked to take a test of about three hours. The Corrigibility track, supported by the Corrigibility Research Fund, emphasizes research relevant to systems that are actually built, attention to corrigibility’s risks as well as possible benefits, and work that does not notably accelerate AI capabilities.

Track availability and funder support are features of this published cycle. They should not be assumed to continue in a later round. The project descriptions can, however, help researchers understand the kinds of questions the organizers considered relevant during this recruitment cycle.

Who the program sought

The official eligibility description is broad in discipline but expects strong research ability. It names mathematics; neuroscience and cognitive science; dynamical systems; physics; philosophy, especially philosophy of science, mind, or ethics; political and economic theory; ecology and evolutionary biology; linguistics; media studies; and the humanities. The list is expressly open-ended. The track descriptions add areas such as theoretical computer science, formal epistemology, software engineering, machine learning, mechanism design, game theory, history, sociology, and law.

The fellowship was aimed at PhD and postdoctoral researchers but accepted substantial research experience regardless of formal credentials. A PhD and published papers were not requirements. Instead, the FAQ asks for evidence of research independence and ability comparable to PhD-level research or equivalent independent work. This allows serious unpublished work, a thesis, a research project, or other clear evidence of independent thinking to matter alongside a conventional academic record.

Prior AI safety expertise was not required, and coding was not mandatory. The program said applicants did not need to bring a finished project idea: project development and mentor matching happened during interviews. It accepted applicants from all countries. Cape Town participation was generally in person; remote attendance was reserved for truly exceptional cases, roughly one person per cohort according to the FAQ. The work was full time and could not be pursued part time or alongside another job or fellowship. The page also emphasizes strong spoken and written English and encourages work on long-term, foundational questions rather than introductory or short-term operational projects.

How applications worked

For the closed winter round, the first stage was a written application. The published materials were a CV or résumé and a personal statement of 600–800 words about the applicant’s research background and motivation; work samples were optional but recommended. The FAQ says the application form included sub-questions and advises answering those directly. These are the requirements recorded for the 2026–2027 cycle, not a live application checklist for a future round. Acceptable samples included papers, unpublished drafts, blog posts, newspaper articles, and similar writing that shows how a person thinks and communicates. Code or a GitHub repository could demonstrate coding ability, but the FAQ cautioned that code alone would not establish research ability.

The later stages involved multiple interviews and project development. Interview conversations explored research interests, shaped possible projects, and supported mentor matching. Applicants were not expected to arrive with a fully specified proposal. After interviews, staff identified possible mentors based on each fellow’s interests and goals. Fellows typically had two or three mentor interviews, then both fellows and mentors submitted preference lists; matching was completed from those rankings. The stated mentoring expectation was at least an hour each week, though the amount and style varied by mentor.

The FAQ describes work tasks as assessments of research-relevant skills, with examples such as writing a proposal, analyzing papers, or critiquing scientific writing. It said no special preparation was needed and that these tasks generally took place in late July, usually within a particular week, with possible case-by-case accommodation. These instructions describe the closed application round and are included for historical context; the organization has not published equivalent instructions for a future cycle.

What to do if interested in a future round

There is no confirmed next-round application deadline or announced set of benefits on the official fellowship page. The linked expression-of-interest form is the appropriate route for registering interest in future fellowships. It asks for contact information, background and research interests, what the person would like to get involved in, and which updates or future roles they want to hear about. It also offers a checkbox for interest in being a fellow at a future PIBBSS Research Fellowship. An expression of interest is not the same as applying to a currently open fellowship.

If a new call is published, candidates should read that announcement from the beginning rather than reuse this cycle’s terms automatically. In particular, confirm the location and in-person requirement, duration, stipend, accommodation and travel support, any research tracks or funders, the exact submission deadline, and the current application form. The past cycle’s CV, statement, and optional work samples provide a useful preparation baseline, but they are not a promise of the next application’s requirements.

A researcher preparing generally can keep a concise CV current, select one or two writing samples that show independent work, and think through how methods from their discipline could inform an AI safety question. Applicants considering an in-person program should also assess whether a full-time stay in Cape Town would be practical. These are sensible preparations based on the closed cycle’s published structure, not additional requirements announced for a future round.

Frequently asked questions

Can I still apply to the 2026–2027 Winter Fellowship? No. The official program page says applications are closed. The application form link remaining on the page does not change that status.

What was the deadline? July 20, 2026. The official timeline gives this date for the Winter Fellowship.

Has a next fellowship cycle been announced? No new application window or deadline is stated. The expression-of-interest form says the organization is preliminarily planning a fellowship every summer, but that is not a confirmed call with dates or terms.

How much support did the closed round provide? The listed package was a $3,000 monthly stipend, a private bedroom, workday office meals, and a return flight to Cape Town. Short-term visa support letters were available. These figures and benefits are not guarantees for a later cycle.

Was a PhD required? No. The FAQ says a PhD and published papers were not required, but applicants needed to show substantial research ability or equivalent independent experience.

Did applicants need prior AI safety or coding experience? No. Neither prior AI safety experience nor coding was required.

Could fellows participate remotely or part time? Remote participation was described as exceptional; the program was primarily in person. Part-time participation was not available.

Where can I register interest? Use the Expression of Interest form linked from the official fellowship page. Check the PIBBSS Fellowship page for any later announcement.

Official sources

The authoritative reference is the PIBBSS Fellowship page from Principles of Intelligence. It confirms that the 2026–2027 Winter applications are closed, gives July 20, 2026 as the deadline, and lists the cycle’s period, stipend, benefits, program structure, tracks, eligibility guidance, application materials, and interview process. The page links to the Principles of Intelligence Expression of Interest form, which invites people to register interest in future opportunities and describes a summer fellowship as a preliminary plan. Neither source announces a confirmed next application date, so this archive preserves the date of the closed cycle and does not invent a replacement.