The Alignment Research Center (ARC) is a non-profit research organization whose mission is to align future machine learning systems with human interests.
A single 2025 grant of $4,553,935 to Model Evaluation and Threat Research Inc marks the clearest recent signal from Alignment Research Center. That award, listed as a program spin-off, sits alongside a grantmaking profile built around technical AI safety work rather than broad philanthropic funding. The center’s stated mission is to align future machine learning systems with human interests, and its active programs show how it approaches that goal: by funding problem-specific contests and proposal competitions tied to ELK, matrix completion, and related algorithmic questions. In the recent grants data, the work is concentrated in research problems that can be framed precisely, evaluated against explicit criteria, and advanced through prizes. The organization also supports individual problem solvers and small research teams through those contests, with prize amounts ranging from $5,000 to $50,000 in its ELK programs and a $5,000 award level in the matrix completion contest. The grant to Model Evaluation and Threat Research Inc suggests a relationship to technical evaluation and threat-oriented research, consistent with the center’s focus on machine learning safety and mechanistic interpretability.
Alignment Research Center’s subject matter is highly technical. In AI alignment and machine learning safety, it has run ELK proposal contests with prizes from $5,000 to $50,000 for training strategies meant to address the eliciting latent knowledge problem. In mechanistic interpretability, those ELK competitions also targeted proposals that could respond to counterexamples or open new directions. The foundation’s matrix completion contest extended this pattern into theoretical computer science, offering $5,000 for solutions or hardness proofs for two specific matrix completion problems. Across these programs, the focus is on algorithmic and mathematical problems that can be judged on technical merit, rather than on service delivery or broad field grants.
The center’s recent grantmaking shows a highly concentrated size profile at the top end: the largest listed award is $4,553,935, while its active prize programs are structured at $5,000 to $50,000, with one contest fixed at $5,000. The organization is a public charity, not a fund that gives to individuals, and it does not make program-related investments. The program descriptions point to prize-funded problem solving and targeted research support, including open submission mechanisms in some contests. The available recent-grants data shows a single recipient entry, so recurring recipient patterns are not visible in this set.
$4.6M
$3.7M
$5.2M
$9.1M
Sign up for a free Kindora account to access AI-generated insights into this funder's giving patterns, decision-makers, and fit signals.
Get Started FreeFree Kindora accounts unlock side-by-side comparisons with foundations that share this funder's focus areas and giving profile.
Get Started FreeSign up free to see how well your nonprofit fits this funder, get an AI-generated pitch, and unlock similar foundations.
The recent grants data places the listed award in Covina, California, while the recipient country distribution is entirely U.S.-based at 100%. The foundation’s active prize programs are described as global or unspecified in geography, indicating that the research contests are not limited to one region in their stated design. In the recent grants list, the visible recipient is in Covina, matching the broad California location of the organization’s headquarters, though the giving pattern itself is not confined to California.
Its active programs focus on AI alignment, mechanistic interpretability, formal verification for machine learning systems, heuristic estimators, and research prizes and contests. The prize programs specifically target ELK proposals and matrix completion problems, which are framed as technical research questions in machine learning safety and theoretical computer science.
The visible 2025 award is $4,553,935. In the active prize programs, ARC offers $5,000 for the matrix completion contest, and ELK proposal prizes range from $5,000 to $50,000, with some intermediate prizes possible.
Yes, through prize competitions and targeted research support. The beneficiary types listed include AI alignment researchers, machine learning theoreticians, individual problem solvers or prize competitors, and small research teams and labs.
For the matrix completion contest, yes: the program accepts unsolicited submissions. The ELK proposal programs include contests where submissions were shared through a Google Doc template, and one ELK program is marked as not accepting unsolicited submissions while another is marked as accepting them.
2025
Source: IRS Form 990-PF, fiscal year 2025.
Most recent grants reported to the IRS.
| Recipient | Location | Amount | Year | Purpose |
|---|---|---|---|---|
| MODEL EVALUATION AND THREAT RESEARCH INC | COVINA, CA | $4,553,935 | 2025 | PROGRAM SPIN-OFF |
MODEL EVALUATION AND THREAT RESEARCH INC
$4,553,935PROGRAM SPIN-OFF