AI Safety Fellowship
Overview
While AI systems become more powerful, our ability to understand, control and regulate these models remains limited.
As a participant in the fellowship, you will partake in weekly discussion centered around our curriculum's core topics. These discussions will be directed by an experienced discussion facilitator. By the end of the semester, we aim to guide fellows to start their own impactful project or join an existing one.
Fellowship alumni have a strong record of continuing to learn and make impactful contributions; many have gone on to become Anthropic fellows, MATS scholars, Astra fellows at Constellation, or Congressional Fellows at the Horizon Institute. Fellowship participants and alumni are also favored for career support from AISI and our affiliates.
We facilitate cohorts every semester, so keep an eye out on our Discord and Website.
Who is this for?
Fellowships are open to anyone interested in learning more about AI safety — we expect both those that are new and those that are familiar with AI safety to benefit.
Applications are live!
The Fall 2026 round of applications is now open. We’re running two tracks this fall covering AI Safety Fundamentals and AI Policy, and we highly encourage you to apply to both! Apply by September 9th, 11:59PM EST.
Tracks
-
We cover recent incidents in the world, fundamental AI safety arguments, and the current threat models and strategy.
Fellows learn to answer questions such as:
How can we scientifically evaluate risks from models?
How can we efficiently implement oversight of models?
How can we detect when model objectives are not truly representative of our goals?
How can we develop mechanistic understandings of model behavior?
-
Understanding the impacts of transformative AI and ensuring that systems are developed, deployed, and regulated responsibly.
Fellows learn to answer questions such as:
How should we regulate frontier AI systems?
How can policymakers predict AI progress?
What impacts will transformative AI have on national security and geopolitics?
How do we ensure the benefits of AI are distributed democratically?
Engagement
Research Opportunities
Join impactful projects alongside AISI researchers aimed at a workshop or arXiv publication. This opportunity is offered to select AI Safety Fellowship graduates and by invitation. Reach out to board@aisi.dev to inquire about this opportunity. Include your resume and a description of your interest in AI safety research. We are coordinating with College of Computing and ML@GT faculty to mentor promising AISI research groups.
Reading Groups
We run a reading group which focuses on new exciting papers in empirical AI safety and alignment research. It caters towards a graduate student audience, though undergraduates with exceptional interest are welcome. We meet every Friday 5-6:30PM in Coda C1215 Midtown. More detail here.
Contact board@aisi.dev if you'd like to start a new group for a topic you're interested in.
Speaker Events
We host presentations by leading researchers and industry professionals who share their insights on AI safety and showcase their latest research. These events, open to both members and the public, offer a unique opportunity to explore emerging topics in the field, engage directly with speakers, and connect with others who share your interest in ensuring AI development benefits humanity. To stay up to date about upcoming speaker events, please join our discord server.
Open Events
Build connections within our AI safety community at these welcoming gatherings. Whether we're sharing pizza over member-led presentations, playing trivia games, diving into journal discussions, or walking through key concepts, these meetings create space for meaningful interaction. Often scheduled alongside speaker events, these casual meetups let you explore AI safety topics while getting to know others who share your passion for responsible AI development. These meetings are open to everyone.
