LONGVIEW PHILANTHROPY

Institutional Archetype

THE ADVISER WITH A DISCRETIONARY FUND — Longview Philanthropy does not hold an endowment. It advises donors who give “over $1 million per year,” recommends grants, and runs pooled funds from which it grants directly. Its home page says: “We support major donors who want to direct their philanthropy to the defining challenges of our time.” Its federal return states the job in one line: “WE EXIST TO MOVE AS MUCH IMPACT-ADJUSTED MONEY AS POSSIBLE TO THOSE WORKING TO REDUCE CATASTROPHIC AND EXISTENTIAL RISKS” (Longview Philanthropy; ProPublica, TY2025 Form 990). In AI it is one of the money lines under the evaluator layer: its TY2025 return pays MATS, the fellowship that supplies the evaluators with researchers, $1,052,000, and pays METR’s legal entity $173,000, and its AI Program Director, Michael Aird, is the secretary of MATS’s board. This profile scores the position, not the people who hold it.

The finding is a seam, not a gate. Longview approves no model and grades no lab. What it holds is a discretionary AI fund whose grantees are not published on the pages read, a program director seated on the board of one grantee, and a donor base the public return does not name. MATS’s own transparency page lists no Longview line, and Longview’s return prints one.

Mandate & Origin

Self-description. “Longview has offices in Washington, D.C., San Francisco, and London, with staff also based in New York.” The grantmaking page: “We focus on risks posed by advanced AI, nuclear weapons, and biological threats—and on ensuring advanced technologies develop in ways that serve the public interest.” Its principles include “Radical impartiality” (“every individual counts equally—including members of future generations”) and “Hits-based giving: Inspired by venture capital, we seek to uncover neglected opportunities with high-reward potential” (Longview Philanthropy; Longview, Grantmaking).

Origin and legal form. The about page’s founder entry for Natalie Cargill: “Natalie founded Longview Philanthropy in 2018.” The contact block and the organizational-structure page: “Longview Philanthropy is the operating name of Longview Philanthropy USA Inc., a 501(c)(3) public charity in the USA (EIN 93-2664730), and its subsidiary Longview Inc. Ltd, a not-for-profit company limited by guarantee in the UK (company number 14444004).” ProPublica lists Longview Philanthropy Usa Inc, Claymont, Delaware, “Tax exemption issued: Feb. 2024,” with returns for the years to December 2024 and December 2025 (Longview, About; Longview, Organizational Structure; ProPublica, Longview Philanthropy Usa Inc).

The funds. Longview’s grantmaking page lists three regranting funds: the Emerging Challenges Fund (“A public fund managed by Longview Philanthropy”), the Frontier AI Fund (“A private philanthropic fund to prepare the world for frontier AI, intended for donors giving over $100,000”) and the Nuclear Weapons Policy Fund; its AI page adds a Digital Minds Fund (“A private philanthropic fund that aims to understand and promote a reasonable and considerate relationship between human civilization and potential digital minds”). The Frontier AI Fund page: “The Frontier AI Fund (FAIF) is a private discretionary grantmaking fund managed by Longview Philanthropy. In its first nine months of operation between December 2024 and September 2025, FAIF supported 18 organizations working on research, engineering, field-building, communications, and advocacy to shape the development and regulation of frontier AI systems.” It offers an “inaugural disbursement report” to “potential donors looking to donate over $100,000”; the report is not on the page (Longview, Grantmaking; Longview, Artificial Intelligence; Longview, Frontier AI Fund).

Who runs it. The about page lists a leadership team of Katie Hearsum (President), Simran Dhaliwal (CEO), Alysha Glenn (Chief Partnerships Officer), Michael Aird (AI Program Director) and Ruby Emerson (Chief of Staff), and an advisory board of Louis Salkind (President, Bright Horizon Foundation), Luke Ding (investor and philanthropist) and Taryn Jensen (Relationship Lead, Giving Pledge). The AI program lists Suryansh Mehta, Zach Freitas-Groff, Ben Stewart, Ajay Karpur, Will Hodgkins, Harrison Gietz, Kendrea Beers, Austin Smith and Luke Marks. Karpur “leads Longview’s grantmaking on AI security” and was before that on “the AI security team at the RAND Center on AI, Security, and Technology, where he co-authored Securing AI Model Weights and advised US government partners on securing frontier AI”; Hodgkins “leads Longview’s grantmaking on technical AI governance, including compute verification and infosecurity.” Aird “leads our AI grantmaking,” joined “from the RAND Center on AI, Security, and Technology, where he was an associate director,” and is “a board member at the MATS Program” (Longview, About; Longview, Ajay Karpur; Longview, Will Hodgkins; Longview, Michael Aird). See Michael Aird.

Governance on the return. The TY2025 Form 990 reports three voting members of the governing body, three of them independent, 18 employees, a conflict-of-interest policy and a whistleblower policy. Schedule O: “DIRECTORS AND OFFICERS ARE REQUIRED TO READ THE CONFLICT OF INTEREST POLICY WHEN JOINING AND TO DISCLOSE ANY CONFLICTS ON A CONFLICT OF INTEREST DISCLOSURE FORM.” It lists Natalie Cargill as President ($98,922 reportable compensation, $27,887 other), Andrew Player as Treasurer and Secretary ($130,554, $45,465), Ben Millwood, Justin Rockefeller and Tyler Whitmer as Directors ($0 each), and Aidan O’Gara (Chief Communications Officer), Carl Alexander Robichaud (Nuclear Weapons Program Director) and Matthew Gentzel (Nuclear Weapons Policy Program Officer) among the highest paid (ProPublica, TY2025 Form 990; ProPublica, Schedule O; ProPublica, Longview Philanthropy Usa Inc).

Funding & Backers

Money in. TY2025 (calendar 2025; submitted 18 September 2026): revenue $21,167,086, of which contributions $20,927,481 (98.9 percent) and investment income $239,605; expenses $15,612,226; total assets $12,681,548, liabilities $389,785, net assets $12,291,763. TY2024: revenue $10,007,719, of which contributions $9,999,060 (99.9 percent); expenses $3,269,897; net assets $6,737,466 (ProPublica, Longview Philanthropy Usa Inc; ProPublica, TY2025 Form 990).

Who the donors are. The public copy of Schedule B prints the contributor as “RESTRICTED” in both years and no name (ProPublica, TY2025 Schedule B). Longview’s pages read name no donor; its about page lists a “Supporters” tab heading and no names in the text served. METR’s about page names “pooled funds such as those of Longview Philanthropy and Effektiv Spenden” among its own supporters; FAR.AI’s donate page says its work has been “endorsed by reputable organizations dedicated to effective philanthropy, such as Founders Pledge, Coefficient Giving (previously Open Philanthropy), and Longview Philanthropy” (METR, About; FAR.AI, Donate). See Good Ventures, Open Philanthropy (Coefficient Giving) and the Survival and Flourishing Fund for the funders that print their lists.

Money out, TY2025. Part I line 13: grants and similar amounts paid, $12,599,230 (TY2024: $1,347,842). The return’s program statement: “GRANTMAKING - GRANTS FROM VARIOUS INTERNAL FUNDS TO HELP FURTHER THE ORGANIZATION’S MISSION,” expenses $13,283,347 including grants of $12,599,230; “ADVISING - OFFER DONOR EDUCATION, GRANT RECOMMENDATIONS, AND GRANT OUTCOME REPORTING TO HIGH-NET-WORTH PHILANTHROPISTS,” expenses $823,953. Schedule I Part II prints fifteen domestic organizations, Part III one individual, and Schedule F Part I prints $6,358,747 to recipients in Europe, which sum to the $12,599,230 on line 13:

(ProPublica, TY2025 Schedule I; ProPublica, TY2025 Schedule F; ProPublica, TY2025 Form 990). TY2024’s grants were smaller and domestic lines were Players Philanthropy Fund ($51,000), Rethink Priorities ($219,000), AI Objectives Institute ($18,843), Pacific Forum International ($50,000) and the Horizon Institute for Public Service ($116,285), with $892,714 in Europe; the five Schedule I lines and the Schedule F total sum to $1,347,842 (ProPublica, TY2024 Schedule I; ProPublica, TY2024 Schedule F).

The money, side by side with the recipients’ own pages. MATS’s transparency page lists 2025 cash gifts from Coefficient Giving (through two intermediaries), Good Ventures Foundation, BERI, the Survival and Flourishing Fund, the Cooperative AI Foundation, Schwab Charitable, Manifund and one individual, and no line labelled Longview; its March 2026 report thanks “Coefficient Giving, the Survival and Flourishing Fund, and Longview Philanthropy, without whose donations we would be unable to run upcoming programs or retain team members essential to this report.” Longview’s TY2025 return prints $1,052,000 to MATS Research Inc (MATS, Transparency; MATS, AI Safety Talent Needs in 2026; ProPublica, TY2025 Schedule I). The lines the transparency page itemizes for 2025 sum to $20,891,404, and $20,011,710 of that is Coefficient Giving and Good Ventures, which leaves $879,694, less than the Longview line; the pages do not say whether the Longview money reached MATS under another name, in another period or on a page not read.

Institutional Voice & Intent

Longview’s public register is the grantmaking essay and the request for proposals, in the first person plural, with the argument made at the level of markets. Its AI page:

• On why philanthropy: “As governments race to adapt and market incentives sideline safety, private markets systematically underinvest in managing AI risks. The dangers of advanced AI affect everyone, but profits are concentrated in a handful of companies, giving them little reason to prioritize societal interests over short-term gains.”
• On what it funds: “Our grantmaking supports cutting-edge technical research, policy development, and field-building efforts that address the most urgent risks—while preserving AI’s extraordinary potential.”
• On the risks it lists: “Misuse by malicious actors,” “Concentration of power,” “Geopolitical instability” and “Catastrophic misalignment” (Longview, Artificial Intelligence).
• On METR, a featured grant: “How would we know if the most powerful AI systems pose catastrophic risks before they’re deployed? METR is building the tools to find out.” The same page features the Center for Human-Compatible AI at UC Berkeley and an interpretability lab at Harvard (Longview, Artificial Intelligence).
• The request for proposals on extreme power concentration (applications closed 2 July 2026): “Longview Philanthropy is looking to fund proposals to better understand and reduce the risk of AI enabling the extreme concentration of power—the possibility that AI will enable a small group of people to gain durable control over the lives of everyone else.” It expects “typical grants to initially be between $100K and $2M/yr” across twelve areas, among them “AI integrity and secret loyalties,” “Frontier company policies against extreme power concentration,” “Oversight of AI in government, including national security” and “Protecting against AI-enabled mass surveillance” (Longview, Request for proposals).

Stated intent: Move “impact-adjusted money” to “those working to reduce catastrophic and existential risks,” in AI through technical research, policy and field-building (Form 990 mission; AI page).

Observed intent: Consistent on the filing: the largest AI-labelled lines are a training pipeline (MATS, NeuroMatch), a pre-release evaluator (METR), a federal-talent fellowship (Horizon), a policy research center (RAND) and a security nonprofit (Foundation for Secure AI), with $6.4 million to Europe unnamed.

Gap: The return prints more than Longview’s pages do. Its AI page features METR and does not give an amount; the return gives $173,000. Its Frontier AI Fund page counts 18 supported organizations and names none; its return names fifteen domestic grantees and none of the European ones. MATS’s transparency page lists no Longview line while MATS’s report thanks Longview, and Longview’s return prints the grant. The gap is structural: the AI Program Director who leads the fund also sits on the board of a grantee whose public donor list does not carry the funder’s name (Longview, Frontier AI Fund; Longview, Michael Aird; MATS, Transparency).

Interest: Longview argues for AI-risk philanthropy in its own pages. Documented, each position beside a term in the record; not a claim about motive and not a verdict that the argument is wrong:

• Philanthropy must fund AI risk because markets will not. Position, in its words, above. What it pays: the case for the pooled funds Longview manages and for the advising it offers donors over $1 million, for which its return reports $0 of program-service revenue. Best case, its own: profits “concentrated in a handful of companies” leave “little reason to prioritize societal interests.” Who else gains: every grantee on Schedule I, among them the MATS pipeline and METR.
• Test dangerous capabilities before release. Position, its description of METR: it “has evaluated GPT-5, Claude 4, and other cutting-edge models before public release—testing for dangerous capabilities and vulnerabilities.” What it pays: the evaluator layer, of which METR is the named featured grant and for which Longview’s return pays $173,000. Best case, Longview’s: no one can know whether the systems “pose catastrophic risks before they’re deployed” without the tools. Who else gains: the labs that use METR’s tools, and METR’s other funders (METR, About).
• Watch for concentrated power in AI, including in government. Position: the RFP’s “Oversight of AI in government, including national security.” What it pays: the grantee field its RFP creates, between $100,000 and $2 million a year per grant; no award is published. Best case, its own wording: “the possibility that AI will enable a small group of people to gain durable control over the lives of everyone else.” Who else gains: the oversight researchers and the people who would be overseen.

Position in the Apparatus

• The AI-labelled grantees on its return, and what each is. METR is the pre-release evaluator profiled at METR; MATS is the pipeline at MATS; the Horizon Institute for Public Service describes itself: “Our Horizon Fellowship trains and places fully-funded fellows at federal agencies, congressional offices, and think tanks” (Horizon Institute for Public Service); RAND is profiled at RAND; Foundation for Secure AI, Rethink Priorities, NeuroMatch, Future Tech Insights and Panoplia Laboratories are listed on the return and have no profile here.
• Seats. Michael Aird is secretary of MATS’s board with Josh Jacobson as president; Ryan Kidd is MATS’s CEO. MATS’s program talent manager’s bio lists “Longview Philanthropy” among the places he built selection systems for (MATS, Team). Aird is also “an affiliate at the Centre for the Governance of AI” (Longview, Michael Aird); see GovAI.
• Beside the larger checkbooks. Coefficient Giving, Good Ventures and the Survival and Flourishing Fund appear in MATS’s list and Longview does not; METR’s about page lists Longview’s pooled funds beside the Survival and Flourishing Fund’s recommendations and Schmidt Sciences. The evaluator-by-evaluator overlap is set out in Evaluator Alliances; see Good Ventures and Halcyon Futures for funders of the same layer at other scales.

Actions & Leadership Choices

CONDUCT: STRUCTURAL — AN ADVISER’S GRANTS TO THE PIPELINE AND THE EVALUATOR, WITH ITS AI PROGRAM DIRECTOR ON THE PIPELINE’S BOARD. No wrongdoing is alleged. The structural point is a loop in plain sight: an adviser raises $21.2 million from donors a public return does not name, grants $12.6 million, puts $1,052,000 into the program that trains the evaluators’ staff and $173,000 into the evaluator it features, and its AI program director holds a fiduciary seat on that program’s board. The funder’s best case is that the seat is on its own staff page, that the return prints every line, that three directors are independent and a conflict policy is in force, and that a grantee thanks it by name in a public report.

The grants rose more than nine-fold in a year. TY2024 grants were $1,347,842 and TY2025 $12,599,230; revenue went from $10,007,719 to $21,167,086. The return does not say which funds the grants came from beyond “VARIOUS INTERNAL FUNDS.”

A government pipeline on the grant list. The TY2025 return pays the Horizon Institute for Public Service $1,600,000 in general support, after $116,285 in TY2024; Horizon places fellows in federal agencies and congressional offices by its own account. Longview’s about page describes “an ongoing partnership with the OECD to support their work on existential risk policy across several directorates, including the Strategic Foresight Unit, the High-Level Risk Forum, the AI Policy Observatory, and the Secretariat on Bio-, Nano-, and Converging Technologies” (Longview, About).

Alliance Map

These run at the same time. One adviser, six years old, moves money to the training pipeline, a pre-release evaluator and a federal-talent fellowship, runs a private AI fund whose grantees it does not list, and seats its AI program director on the pipeline’s board; the pipeline’s donor page does not carry the funder’s name while its report does. Longview, MATS and METR each disclose their own side of the tie. The overlap is the finding.

Reach Assessment

Institutional: $12,599,230 in grants in TY2025 against $1,347,842 in TY2024; $21,167,086 of revenue; 18 employees; a private AI fund with 18 supported organizations in nine months; a named grant to the pipeline that trains the evaluators and a named grant to the evaluator it features; a partnership with the OECD’s policy units. Memetic: Its pages, invite-only convenings and testimonials from named donors and officials; no research output or policy paper is published under its name on the pages read.

Civilizational: For scale, the Survival and Flourishing Fund organized about $216MM since 2019 and Coefficient Giving says its technical AI safety, security and fieldbuilding commitments were $351 million in 2025; Longview’s TY2025 grants are about a fifth of the Survival and Flourishing Fund’s 2026 round of $63.13MM (Survival and Flourishing Fund; Coefficient Giving, Emily Oehlsen, Sep 9 2026). Longview is a mid-sized allocator whose AI lines run through the same few grantees as the larger ones.

Score note. Attack 5, defense 4 and hp 4 on first-hand receipts: the TY2025 return’s revenue, grants and Schedule I, Schedule B and governance answers, and Longview’s own fund and staff pages. Peers: Halcyon Futures (5/4/3, hazard 40), a seed funder with $611,455 in grants and unnamed donors, level on reach and insulation and a point lower on durability because its FY2024 balance sheet is cash and receivables with no standing funds; Survival and Flourishing Fund (6/8/6, hazard 67), which allocates $63.13MM in a round and prints its funders; Good Ventures (7/9/8, hazard 80), whose single donor’s endowment pays the same layer directly. At person level, Michael Aird (5/4/4, hazard 43.3) scores level with the charity, because his post is the AI program of the charity and his board seat matches its reach. Hazard 43.3 sits just above Halcyon and well below SFF: reach one lower, insulation four lower, entrenchment two lower.

Score basis. Atk 5 is reach: $12,599,230 in TY2025 grants and Schedule I lines to MATS, METR, RAND and the Horizon Institute, a private AI fund that supported 18 organizations in nine months, and a program director on a grantee’s board, with no statute or gate obliging anyone to use a grantee’s output (ProPublica, TY2025 Schedule I; Longview, Frontier AI Fund). Def 4 is funding insulation: 98.9 percent of $21,167,086 is contributions from contributors the public Schedule B prints as “RESTRICTED,” while three of three voting directors are independent (ProPublica, TY2025 Schedule B; ProPublica, TY2025 Form 990). Hp 4 is entrenchment: founded 2018, exempt in the US since February 2024, 18 employees, $12,291,763 in net assets, standing funds and an invite-only donor network; none is held in place by statute or contract (ProPublica, Longview Philanthropy Usa Inc).

Sources: Longview Philanthropy; Longview, About; Longview, Grantmaking; Longview, Artificial Intelligence; Longview, Frontier AI Fund; Longview, Organizational Structure; Longview, Request for proposals; Longview, Michael Aird; Longview, Ajay Karpur; Longview, Will Hodgkins; ProPublica, Longview Philanthropy Usa Inc; ProPublica, TY2025 Form 990; ProPublica, TY2025 Schedule B; ProPublica, TY2025 Schedule F; ProPublica, TY2025 Schedule I; ProPublica, TY2025 Schedule O; ProPublica, TY2024 Schedule F; ProPublica, TY2024 Schedule I; METR, About; FAR.AI, Donate; MATS, Transparency; MATS, Team; MATS, AI Safety Talent Needs in 2026; Horizon Institute for Public Service; Survival and Flourishing Fund; Coefficient Giving, Emily Oehlsen (Sep 9 2026)