Supervised Program for Alignment Research

SPAR (Supervised Program for Alignment Research) is a part-time, remote research fellowship that matches aspiring AI safety and policy researchers with experienced mentors for structured, 3-month research collaborations. Launched in Spring 2023 through the Berkeley AI Safety Initiative for Students (BASIS), the program was originally co-organized with Stanford AI Safety and Georgia Tech AI Safety. It drew inspiration from similar mentorship initiatives at Stanford AI Alignment and the OxAI Safety Hub. From its university-focused roots, SPAR expanded to accept participants at all career stages — undergraduates, graduate students, and working professionals alike — and grew to become one of the largest AI safety research fellowships in the world. The Fall 2025 round featured over 300 mentees across 55 countries working on 80+ projects, and the Spring 2026 round launched with 130+ projects, described by the organizers as the largest round of any AI safety research fellowship to date. In October 2024, SPAR became a core program of Kairos AI Project, Inc., a newly formed US 501(c)(3) nonprofit focused on accelerating the AI safety talent pipeline. Alongside the Pathfinder Fellowship (which supports AI safety university groups globally), SPAR constitutes one of Kairos's two flagship programs. Kairos is co-directed by Agustín Covarrubias and Neav Topaz, with Rebecca Baron as Founding Generalist. Research areas offered through SPAR include technical alignment, interpretability, AI governance and policy, AI security, and biosecurity. The program covers project expenses such as compute, API access, and software. Most projects result in published papers or preprints, and SPAR research has been accepted at top venues including ICML and NeurIPS, and has been covered by TIME. Demo Day events feature prizes totaling $7,000 and direct connections to hiring organizations in the AI safety ecosystem. Kairos is funded by Coefficient Giving, and Open Philanthropy has recommended grants to Kairos in support of its AI safety talent development work.