FAR.AI / FAR.Futures field-building grant

FAR.AI (Frontier Alignment Research) is an AI safety research nonprofit founded in July 2022 by Adam Gleave and Karl Krueger, headquartered in downtown Berkeley. Open Philanthropy recommended $6.65M over two years specifically to scale FAR.Futures, the organization's field-building arm. FAR.AI separately announced over $30M in multi-funder commitments in 2025 across its research and field-building lines combined.

FAR.Futures runs the Alignment Workshop series, which has now hosted events in San Diego, Singapore, Vienna, New Orleans, San Francisco, and London. The grant supports expanding the series from two to three events annually and extending into Asia and the Global South, plus specialized technical workshops, hackathons, and consensus-building forums. A February 2026 deception-focused workshop in San Francisco gathered 25 researchers to specifically work on scheming detection and chain-of-thought monitorability.

FAR.Labs, the Berkeley co-working hub, hosts a weekly FAR.AI Seminar featuring presentations from external researchers, streamed remotely. The fellowship and internship pipeline supports 20+ fellows annually with mentorship and research opportunities. A separate targeted grants program identifies and scales third-party trustworthy-AI research.

FAR.AI's own research output centers on adversarial robustness and scheming. Their empirical work on robustness scaling found that adversarial training is orders of magnitude more compute-efficient than model-size scaling for robustness gains, but that the offense-defense balance still favors offense in absolute and relative terms. Their Safety Cases line studies vulnerabilities in safety features and develops methodologies to quantify risk levels and scaling laws for harm. FAR.AI is also a December 2025 AISF grantee for "Quantifying the Safety-Adversary Gap in Large Language Models."

The Open Philanthropy URL for the field-building grant now redirects to coefficientgiving.org following the funder's late-2025 rebrand. Among AI safety nonprofits, FAR.AI is structurally distinctive: it runs convening, research, and infrastructure under one roof, with the Berkeley physical hub as connective tissue. The $6.65M grant funds the convening and outreach layer specifically, separate from the research grants the organization receives elsewhere.