Cambridge Boston Alignment Initiative

The Cambridge Boston Alignment Initiative (CBAI) is a Cambridge, Massachusetts 501(c)(3) public charity (EIN 92-1463153) that serves as the institutional and fiscal backbone of the Boston-area AI-safety student community. Founded in 2022 and built around two student organisations — MIT AI Alignment (MAIA) at MIT and the Harvard AI Safety Team (AISST) at Harvard — CBAI provides fiscal sponsorship, programming infrastructure, and direct research support to accelerate the pipeline of technically skilled students into AI-safety careers. With nearly $2 million in annual revenue as of its 2024 fiscal year, it is among the better-resourced student-community infrastructure organisations in the AI-safety field.

Founding and institutional origin

CBAI emerged from the same founding moment as MAIA and AISST. In fall 2022, a founding team that included Trevor Levin and Sam Marks transformed an informal MIT–Harvard reading group into two distinct student organisations — MAIA at MIT and AISST at Harvard — while simultaneously establishing CBAI as the 501(c)(3) umbrella that could hold the financial and legal infrastructure neither student group could carry alone. CBAI's 501(c)(3) status was recognised by the IRS effective November 2023. Kuhan Jeyapragasan, who co-organised MAIA's early retreats, serves as President; Levin as Secretary; and Marks as a board member. The January 2023 MLAB-inspired ML bootcamp — one of the first concrete CBAI outputs — foreshadowed what would become the CAMBRIA programme.

Programs

CAMBRIA. CBAI's flagship programme is the Cambridge Bootcamp for Research in Interpretability and Alignment (CAMBRIA) — a three-week in-person ML upskilling programme built on the ARENA curriculum. Cohorts cover transformers, deep reinforcement learning, and mechanistic interpretability. CAMBRIA is open beyond MIT and Harvard: any applicant with Python proficiency, multivariable calculus, and linear algebra can apply, and CBAI provides housing, meals, 24/7 office access, and travel support. In 2026, three cohorts run across two cities: two at Harvard Square (May–June, August) and one in Manhattan, hosted by Collider (July). The programme descends directly from the MLAB-inspired bootcamp the founding team planned for January 2023 — an early articulation of CBAI's intent to serve technically capable students who had outgrown the introductory fellowship.

AIxBiosecurity Summer Fellowship. CBAI runs a summer fellowship focused on the intersection of AI and biosecurity risks — a thematic niche reflecting the founding community's concern with the most catastrophic risk scenarios at the frontier of AI capability.

Summer Research Fellowship. A direct research-placement fellowship providing support for students engaged in AI-safety research projects over the summer period.

Scale and operations

CBAI's FY2024 Form 990 shows revenue of $1,968,408, nearly all from contributions, and total assets of $1,361,474. The FY2023 figure was $1,497,763 — growth of approximately 30% in one year. The organisation employs a professional staff team including a Director of Programs (Emre Yavuz), Director of Operations (Maite Abadia-Manthei), three Research Managers (Alex Semendinger, Claudio Mayrink Verdun, Dmitrii Troitskii), and an Operations Associate. This staffing level — comparable to a small foundation — supports the two student groups and the programmes CBAI runs directly.

Coefficient Giving (formerly Open Philanthropy) is CBAI's primary documented institutional funder, with multiple confirmed grants spanning 2023 to 2025: $1,170,000 for AISST and MAIA operating costs, $550,000 for Harvard AI safety education and outreach, $387,741 for AI safety field-building, and $350,000 for the 2025 Summer Research Fellowship. The scale of this support explains how CBAI built a professional staff team within two years of founding.

Place in the movement

CBAI occupies an unusual structural position in the Boston AI-safety ecosystem: it is not itself a research group or a public-advocacy organisation, but the institutional layer that makes the student pipeline durable. Without fiscal sponsorship, MAIA and AISST would have no stable mechanism to receive funding, employ staff, or run multi-year programming — the student-leadership churn inherent in university groups would erode continuity. CBAI's fiscal-sponsor model solves this problem, providing a stable 501(c)(3) shell while preserving the student-led character of the groups it sponsors.

The CAMBRIA programme extends CBAI's reach beyond the MIT–Harvard student community. By running open bootcamps and partnering with New York-based venues like Collider, CBAI is beginning to function as a regional infrastructure node for AI-safety upskilling on the US East Coast. The AIxBiosecurity focus adds a thematic dimension: unlike the more general alignment-research orientation of MAIA and AISST, the biosecurity fellowship tracks the specific risk intersection that has drawn increasing attention from technical safety researchers.