Existential Risk Laboratory Fellowship Program (2025)

Existential Risk Laboratory Fellowship Program (2025) is sponsored by Open Philanthropy. Funds undergraduate and graduate students working on independent research projects in AI and other existential risk areas at the University of Chicago.

Summer Research Fellowship 2026 | XLab The University of Chicago SUMMER RESEARCH FELLOWSHIP 2026 The Existential Risk Laboratory (XLab) at the University of Chicago is accepting applications for the 2026 Summer Research Fellowship (SRF), a 10-week, in-person research program for students and early-career researchers working on AI safety and security, nuclear security, or the confluence of these issues.

Unlike more established fields, the field of AI safety and security has not yet reached a consensus on what the foundational problems are . The rate at which advances occur together with the youth of the field means that important questions and approaches have likely been overlooked.

At the same time, the literature is shallow enough that dedicated junior researchers can reach the intellectual frontier with just months of dedicated work. The confluence of AI and nuclear security is arguably even more neglected — vanishingly few researchers have fluency in both domains, leaving critical questions almost entirely unexamined.

And within nuclear security itself, the nonproliferation field has traditionally underinvested in right-of-boom issues such as nuclear winter and civil defense, relying on Cold War–era assumptions about deterrence and escalation dynamics that are increasingly outdated.

The Summer Research Fellowship aims to give early-career researchers the time, resources, and intellectual community they need to identify consequential but overlooked questions across all of these areas. Our program is designed to enable fellows to develop their own research agendas.

Each fellow scopes their own research question, justifies its importance, identifies appropriate methods, and delivers a significant written product: a journal article, workshop submission, white paper, or equivalent. They are advised by a domain expert mentor, but the intellectual direction of the project is theirs. This is not intended to be a research assistantship.

It is intended as practice for leading your own research program. We’re looking for applicants committed to careers in AI safety or nuclear security who just need the time, resources, and community to develop their ideas. If that describes you, we want to hear from you.

APPLICATION FORM WILL REMAIN OPEN UNTIL MARCH 18 APPLICATIONS ARE DUE MARCH 15 AT 11:59 PM (YOUR TIMEZONE) . Please don’t spend more than 3 hours on the application. Applicants with existing research proposals should be able to finish the application in ~1/2 hour.

The program will run for 10 weeks from June 15th to August 22 , in-person on the University of Chicago campus. Stipend: All fellows receive a $10,000 stipend, University of Chicago dorm housing, a full dining hall meal plan, and funds for weekday lunches. Compute: Technical fellows receive $4,000 in compute and API credit support.

Mentorship: Fellows have regular one-on-one check-ins with a research manager throughout the program. Admitted fellows, with program staff assistance, will be responsible for identifying and reaching out to appropriate domain experts to advise their work. Workspace: Dedicated coworking space on the University of Chicago campus.

Programming: Workshops on research methodology, including forecasting, open-source intelligence, and research process skills, alongside Q&A sessions with policy practitioners and active researchers on open questions and career paths. Community: A cohort of 15–20 peers working on related problems. Social events, peer review sessions, and shared living spaces foster lasting professional relationships.

Continued involvement: Fellows may have the opportunity to continue their work with XLab as affiliate researchers depending on their contributions during the summer. Commitment: Fellows are expected to make the program their primary commitment for the duration of the fellowship. Fellows should not be concurrently working substantial hours in other labs or internships or taking full course loads unless directly relevant to their project.

The SRF is a demanding program.

Our ideal applicant has demonstrated commitment to a career in AI safety, security, and governance or nuclear security substantive familiarity with the field and its open questions shown an ability to work and think independently This program is particularly valuable for people who have built real knowledge of the field but haven’t yet had the opportunity to develop a substantial, sole- or lead-authored research product or explore their interests full-time.

No technical background is required, but we expect fellows pursuing technical projects to have completed ARENA or similar material. If you’re on the fence, we encourage you to apply. Some of our strongest fellows were deeply uncertain about whether they were ready or whether they were even the kind of person who could contribute.

If you have a serious interest in these issues and have ideas you want to develop, we want to see your application. Note : We are only able to accept fellows with US work authorization: US citizens, permanent residents, or international students currently studying in the United States (typically on a current F-1 visa). We are unable to sponsor visas.

We strongly encourage applicants with US work authorization who are studying at international universities to explicitly note this in their application. Since 2022, the SRF has produced 60 alumni. Fellows have gone on to positions at OpenAI, RAND, Google DeepMind, MATS, SecureBio, the Federation of American Scientists, the Institute for Progress, MILA, the intelligence community, congressional offices, and leading PhD programs.

Response Inconsistency of Large Language Models in High-Stakes Military Decision Making Deterrence and Deescalation: U.S. Strategies for a Nuclear-Capable Iran Certified Safe: A Schematic for Approval Regulation of Frontier AI A California Effect for Artificial Intelligence Rethinking Priorities for Frontier AI Security Position: Why AI R&D Benchmarks Need to Measure Research Judgment Separately Open-source Language Models: Identifying Contributing Factors and Future Trends Forbidden Facts: An Investigation of Competing Objectives in Llama-2 Global Governance of High-Risk Artificial Intelligence Our program focuses on AI strategy, security, and governance and nuclear weapon security.

But, we see these areas not as separate issues to be siloed but facets of the same problem. A single research question can cascade across domains. The satellite imagery analysis techniques that track compute buildups for AI governance also inform long-standing questions about arsenal survivability.

If vision models can identify data centers from orbit, what does that imply for mobile missile launchers that once relied on concealment? Work on adversarial materials that obscure targets from vision models informs both whether lethal autonomous weapons can be systematically defeated and whether high-value strategic assets can be hidden.

As AI growth intensifies great-power competition, nuclear-armed states may face use-it-or-lose-it pressures if they believe AI-enabled technology will render their arsenals vulnerable, or may threaten use to defend interests they consider existential.

We are especially interested in encouraging applications for projects that employ analytic methods such as open-source intelligence, forecasting, strategic net assessment, or scenario analysis. We have historically received fewer applications proposing to take these approaches in the past. However, we’re just as excited to review proposals for technical governance, AI security, cybersecurity, and technical applications of AI technology.

“I have loved every minute of x-lab, particularly the opportunity to meet so many like-minded but uniquely smart people working on subject areas different from mine. I’ve learned to push myself more on questioning my own assumptions, work with teams, and produce a tailored final deliverable. ” “It was quite different from what I anticipated but in a better way.

I anticipated the program to be more about just me and my mentor working on a research project, but there was a good amount of interaction with my cohort that I didn’t expect. I met many great people and also learned a lot of things I didn’t know from the weekly workshop and group meeting (such as unfamiliar areas like nuclear and biotech threats, and different perspectives in AI policy/safety).

” “I didn’t anticipate having so much freedom. While it was intimidating at first, it was definitely the challenge I needed. I feel a lot more confident in my ability to commit to and work on an intellectual pursuit.

” “Going into this program, I had pretty low self-confidence of my capacity to do a technical project and was fully planning to just write a governance report. This program has flipped those expectations, where I better understand where my core weaknesses such that I know that I can initiate and execute a research project decently and also know where I need to build my next skills to have a higher impact.

” “Much better socially than I expected, living in the dorms was great. Being able to co-work with people and talk to others working on technical AI projects was super useful, so much better than counterfactual of just trying to skill up on my own. ” “I was, however, thrilled and pleasantly surprised by how useful it was having conversations with other fellows and being immersed in the cohort.

Without these countless conversations, I would not have developed the foundational understanding of AI that I needed to write my research report. ” “Generally speaking, it was a beautiful experience. I grew an amount that is hard to think, both technically and as a person.

The pressure was high as it was not just ‘helping out,’ rather doing my own project, so responsibility for guiding the research is on me. ” “I am so grateful for XLab because had it not been for this summer, I don’t think I would have had a vision for future research projects and the ability to even know that I could do it myself without any professors. ” “i think the main benefits for me were: 1.

testing fit in individual research — the research is just a very different experience from anything i’ve ever done (more intensive and more independent), so i think that taught me a lot about how i like to work and what systems work better for me 2.

meeting all the other fellows — i previously didn’t really know any friends interested in x-risk or even ai risk because i was so new to it, so meeting people here was really valuable especially since some of them have spent so much time in the field and are just great resources”