GOOD VENTURES

Institutional Archetype

THE ENDOWMENT BEHIND THE ADVISER — Good Ventures is the private foundation of Cari Tuna and Dustin Moskovitz, and the payer on the record: its Form 990-PF is the document on which the grants that Open Philanthropy (Coefficient Giving) recommends show up as money leaving an account. It has no foundation staff to speak of; it says so. The advisor chooses, the foundation pays, the two founders sit on the foundation’s board and on the board that governs the advisor. In the AI layer the result is that the same dollars carry three names depending on which page is read: Good Ventures on the tax return, Coefficient Giving on the grant record, Moskovitz on the donor line and on the fund (the Survival and Flourishing Fund) that names him as a funder in 2026. This profile scores the position, not the donors, the advisor or the grantees.

The finding is arithmetic. In the tax year to 30 June 2025 the foundation took in $1,395,845,381 in contributions and paid out $306,706,819 in grants, 4.6 times as much in as out, while the stated end state is to give the fortune away. Of what it paid, $55,341,516 sits on lines whose printed purpose names AI, large language models, machine learning or alignment, and the recipients include four of the evaluators and pipeline organizations on the evaluator alliances map.

Mandate & Origin

Self-description. The home page: “Our mission is to help humanity thrive,” and “Good Ventures is a philanthropic foundation created by Cari Tuna and Dustin Moskovitz. Our goal is to take the fullest possible advantage of the opportunity we have to help others.” Its method, in its words: it chooses causes “using three criteria: importance, neglectedness, and tractability” (Good Ventures). The governance page: “Good Ventures Foundation is a 501(c)3 private foundation that makes grants and is associated with other entities that make grants and investments in support of our mission” (Good Ventures, Governance).

Origin. The founders’ letter: “When we founded Good Ventures in 2011, we knew we wanted to give away the vast majority of our wealth.” It credits an early conversation with Holden Karnofsky of GiveWell and says it launched “a project, now called Coefficient Giving,” and that “With support from the Coefficient team, Good Ventures has distributed more than $7 billion to address immediate challenges like malaria and lead poisoning, and to prepare for emerging threats like deadly pandemics and increasingly powerful AI systems” (Good Ventures, Our Story). The Our Giving page prints “15+ years of giving” and “$7B distributed to grantees.” ProPublica’s record for the legal entity: Good Ventures Foundation, 314 Lytton Ave, Palo Alto, EIN 46-1008520, IRS ruling date March 2013, private foundation (ProPublica, Good Ventures Foundation). See Holden Karnofsky.

How it works. The home page: “Instead of building a big foundation staff, we work closely with these trusted partners to maximize the impact of our giving.” And of its advisor: “Coefficient Giving (formerly Open Philanthropy) is our primary philanthropic advisor. The Coefficient team helps us choose causes, make grants, track results and pool resources with other donors” (Good Ventures). The return agrees: no officer is compensated, no employee is paid more than $50,000, and the largest line among its paid contractors is “Philanthropic Services Corp,” for “Grant administration,” at $29,964,000 (ProPublica, Good Ventures FY2024 Form 990-PF).

Who runs it. The governance page lists “Cari Tuna, Chair,” “Dustin Moskovitz, Secretary and Treasurer,” and “Divesh Makan, Partner, ICONIQ Capital.” The 990-PF lists the same three as directors, Tuna as President and Board Chair, and Tom Van Loben Sels as Assistant Secretary, each at one hour a week and $0 (Good Ventures, Governance; ProPublica, Good Ventures FY2024 Form 990-PF). The advisor’s own pages state the reverse seat: Coefficient Giving LLC “is governed by a Board of Managers currently consisting of Cari Tuna (Chair of the Board), Dustin Moskovitz, Elie Hassenfeld, and Alexander Berger,” so two of its four managers are two of the foundation’s three directors (Coefficient Giving, Governance).

What the money is for, by the foundation’s own list. Its Our Giving page names seven focus areas, among them Global Health and Development, Biosecurity and Pandemic Preparedness, Farm Animal Welfare and “Navigating Transformative AI,” defined as “Helping humanity prepare for a safe transition to advanced AI, with a focus on preventing worst-case outcomes through research and policy work.” The AI examples it names are FAR AI and BlueDot Impact; under biosecurity it names “SecureBio’s Nucleic Acid Observatory uses computational approaches to detect biological threats in wastewater” (Good Ventures, Our Giving).

Funding & Backers

Money in. The return for the tax year 1 July 2024 to 30 June 2025 (filed 15 May 2026) reports total revenue of $2,597,571,672: contributions of $1,395,845,381, dividends and interest of $71,628,332, a net gain on the sale of assets of $846,587,145 and other income of $283,510,814. Total assets at the end of the year were $10,107,955,038, liabilities $176,270,767 and net assets $9,931,684,271; the return for the year to June 2023 reported assets of $5,863,998,880 (ProPublica, Good Ventures Foundation; ProPublica, Good Ventures FY2024 Form 990-PF).

Who gave it. Schedule B, Part I, lists one contributor: “Dustin A Moskovitz Remainder Interest Trust,” at 314 Lytton Ave, $1,395,695,354, noncash, “Publicly traded securities,” received 30 June 2025. That is 99.99 percent of the year’s contributions. The same return lists Cari Tuna and Dustin Moskovitz as the managers who contributed more than 2 percent of the foundation’s total contributions (ProPublica, Schedule B; ProPublica, Good Ventures FY2024 Form 990-PF). Coefficient Giving’s own account is the same at a larger scale: “he and Cari are Coefficient Giving’s main funders” (Coefficient Giving, Dustin Moskovitz).

Money out. Part XIV lists 566 grants paid during the year, totalling $306,706,819 (line 3a), and 94 more “approved for future payment,” totalling $87,998,087 (line 3b). Part I prints “Contributions, gifts, grants paid” of $293,536,125 on the books and $306,706,819 on the cash basis, and total expenses and disbursements of $369,108,718. The return’s distributable amount is $393,790,093 and its qualifying distributions are $337,695,777; Part XII carries forward excess distributions of $317,576,940 from 2021, 2022 and 2023 (Parts X, XI and XII). Grants paid equal about 3.0 percent of year-end assets, and contributions received were 4.6 times grants paid (ProPublica, Good Ventures FY2024 Form 990-PF).

Paid contractors. Part VII lists the five highest-paid independent contractors: Philanthropic Services Corp (grant administration, $29,964,000), Value Aligned Research Advisors LLC (investment management, $5,264,377), Parametric Portfolio Associates LLC (investment management, $1,316,938), Center for Welfare Metrics (Sao Paulo; project consulting, $416,667) and Sun Life Capital Management (US) LLC (investment management, $376,910) (ProPublica, Good Ventures FY2024 Form 990-PF).

The AI-related lines. Counting the paid lines whose printed purpose or recipient text names AI, large language models, machine learning, alignment, FrontierMath, artificial intelligence or interpretability: 107 lines, $55,341,516, 18.0 percent of grants paid, of which $792,362 went to individuals. Among the approved-for-future lines, 19 such lines total $16,764,502, 19.1 percent. The recipients on the evaluator and pipeline map, as printed:

(ProPublica, Good Ventures FY2024 Form 990-PF). Approved for future payment, the same return lists SecureBio Inc at $2,600,000 for general support, the Hebrew University of Jerusalem at $2,725,000 for a “Governance of AI Lab,” the Regents of the University of California, Berkeley at $3,390,000 to “evaluate LLM capabilities on cybersecurity defense benchmarks,” and Stanford at $1,958,000 for a “benchmark for evaluating LLM cybersecurity risks.” The 566 paid lines carry no recipient named METR, Apollo Research, Redwood Research, Palisade Research, Transluce, Sentinel Bio, Halcyon Futures, Fathom or Ergo Impact, and none named Coefficient Giving or Open Philanthropy as a recipient; 21 individual lines are labelled “Open Philanthropy Scholarship Program” ($1,183,420) and 12 “AI Fellows Program” ($649,176).

Same money, three names. Good Ventures does not say which grants it pays; its advisor does not say which donor funds which grant. Where an amount appears on both records the match is stated and nothing more is inferred.

The advisor’s own description of the relationship: “deploy billions of dollars from Good Ventures, our main funder,” and “Good Ventures (our largest funding partner)” (Coefficient Giving, Our Progress in 2024; Coefficient Giving, Oct 2 2025). It also says “Good Ventures represents a tiny fraction of what’s possible in philanthropy today,” and that in 2024 it “directed over $100 million to high-impact causes from donors besides Good Ventures” (Coefficient Giving, Nov 18 2025).

Institutional Voice & Intent

Good Ventures’ public register is the founders’ letter and a list of partners: short pages, a mission sentence, a press page of other people’s headlines.

• On the arrangement: “Everything we do, we do in partnership. Instead of building a big foundation staff, we work closely with these trusted partners to maximize the impact of our giving” (Good Ventures).
• On the end state: “we’ll keep showing up with the same intention — to take the fullest possible advantage of the opportunity we have to help others — until we’ve moved the last dollar out the door” (Good Ventures, Our Story).
• On the pace, in the advisor’s words: “Cari and Dustin have committed to spending down the vast majority of their wealth within their lifetimes” (Coefficient Giving, Nov 18 2025).
• Its press page lists, among others, an Associated Press piece (18 November 2025) headlined “Billionaire Cari Tuna on why the organization she started will offer free advice to other donors,” and a Financial Times piece (26 May 2026) on a $140 million fund “backed by donors including former Facebook and Asana co-founder Dustin Moskovitz and his wife Cari Tuna” (Good Ventures, In the News).

Stated intent: Give the fortune away within the founders’ lifetimes, to causes chosen on importance, neglectedness and tractability, through an advisor instead of a staff.

Observed intent: Pay what the advisor recommends and add donors to the advisor’s pooled funds. In the tax year to 30 June 2025, the foundation paid $306.7 million against $1.4 billion received, and a fifth of the lines it paid name AI.

Gap: The foundation’s return is more informative than most funders’ pages: it prints every recipient, amount and purpose, names its one contributor, and itemizes its contractors. What the record does not show is the grant decision itself: the foundation’s pages say the advisor “helps us choose causes, make grants,” the advisor’s pages say it recommends, and no page cited says where the choice is made when the two boards share two names. The gap is structural: the checkbook, the staff and the board seats sit in two entities with overlapping directors, and each entity says so.

Interest: Good Ventures argues no AI-policy position on its own pages. The positions it does state concern how it gives. What each would pay, from its pages, its return and its advisor’s pages:

• Give through an advisor, not a staff. Position, in the foundation’s words: “Instead of building a big foundation staff, we work closely with these trusted partners to maximize the impact of our giving.” What it pays: the foundation’s return reports no compensated officer and no employee over $50,000; the home page says the Coefficient team “helps us choose causes, make grants, track results and pool resources with other donors,” and Coefficient Giving’s board of managers includes two of the foundation’s three directors; for the advisor, the position makes the foundation its “largest funding partner” (Coefficient Giving, Oct 2 2025). Best case, the foundation’s own: the arrangement exists “to maximize the impact of our giving.” Who else gains: the advisor’s other donors, to whom it offers “research on cause areas and specific giving opportunities, strategic guidance on how to maximize impact, and operational support” (Coefficient Giving, About).
• Bring other donors into the same funds. Position, the advisor’s: “Additional funders can help build a more effective coalition behind AI safety and security; back areas and organizations that Good Ventures (our largest funding partner) is not well-positioned to support; and increase the total amount of funding in this space,” and “we are typically able to recommend giving opportunities to external funders that are 2-5x as cost-effective as Good Ventures’ marginal AI safety funding” (Coefficient Giving, Oct 2 2025). What it pays: the grantees on both records, which gain a second funder, and the foundation, whose share of any one grantee’s budget falls. The same post gives an example of a scope limit on the foundation’s side: “some conservative groups that might do influential and helpful work on AI policy may not be a good fit for Good Ventures.” Best case, the advisor’s: “Money is less fungible in policy work than in other causes (like delivering direct services) because a group’s funding sources impact its positioning and influence in different ways.” Who else gains: the AI Safety Tactical Opportunities Fund, which the post names as an independent funder, and any donor who wants a vetted list.
• Spend the principal down. Position: “until we’ve moved the last dollar out the door.” What it pays: grantees first; on the return, the pace is $306.7 million of grants against $1.4 billion of contributions in the tax year to 30 June 2025, so the stated end state and the annual flow are on different timescales. Best case, the founders’: “we knew we wanted to give away the vast majority of our wealth.” Who else gains: every recipient on the lists above, and the advisor, which says it will “expect our partnership to grow even more in the coming years” (Coefficient Giving, Nov 18 2025).

Interest is a documented fact about each position, not a claim about motive, and not a verdict that the argument is wrong. Each best case is stated at the strength its author states it.

Position in the Apparatus

• The checkbook behind the advisor. The advisor’s rename post calls it “our founding partnership with Good Ventures, the foundation of Cari Tuna and Dustin Moskovitz” (Coefficient Giving, Nov 18 2025). The AI fund the advisor runs, with more than 570 grants, lists Good Ventures as its partner, and the advisor’s President wrote on 9 September 2026 that its technical AI safety, security and fieldbuilding commitments were $168 million in 2024 and $351 million in 2025 and are on track for over $1 billion in 2026 (Coefficient Giving, Navigating Transformative AI; Emily Oehlsen). The advisor says Good Ventures and it “are currently a concentrated share of AI safety philanthropic funding” (Coefficient Giving, Oct 2 2025). See Open Philanthropy (Coefficient Giving).
• Four doors into one evaluator. SecureBio, a member of the AI Evaluator Forum, is paid by Good Ventures directly ($1,630,000 in the tax year to 30 June 2025, $2,600,000 more approved), by Coefficient Giving (grants listed from $1,420,937 in March 2023 to $4,000,000 in April 2025), by the Survival and Flourishing Fund (SFF-2026: $927,000, of which Moskovitz is the funder of $896,000) and by Sentinel Bio ($600,000, June 2025, on Sentinel’s own page), and says it is “primarily funded by Coefficient Giving, the Survival and Flourishing Fund, and Sentinel Bio” (SecureBio; Sentinel Bio, SecureBio grant; SFF-2026). See Sentinel Bio.
• A second Forum member. RAND, also a Forum member, is on the return at $5,166,667 for general support. Coefficient Giving’s records show “a gift of $2.5 million to RAND Corporation, to be spent at the discretion of RAND president and CEO Jason Matheny” and “a gift of $10 million over three years to RAND in support of its work on Canary,” and Sentinel Bio’s own page records a gift of $756,000 “to be spent at the discretion of RAND president and CEO Jason Matheny” (ProPublica, Good Ventures FY2024 Form 990-PF; Coefficient Giving; Sentinel Bio, RAND grant). See RAND Corporation.
• The talent pipeline, paid directly. MATS ($4,558,450), AI Safety Support ($2,321,550), Safe AI London for ARENA’s bootcamps ($3,755,278 across five lines), ML4Good ($954,911), BlueDot Impact ($300,000) and the “AI Fellows Program” scholarships are lines on the return. MATS’s own page reports $20,011,710 of its $20,891,404 in 2025 cash gifts from Coefficient Giving and Good Ventures Foundation (MATS, Transparency). See MATS.
• An evaluator’s legal entity. The return lists Pattern Labs Tech Inc, Wilmington, Delaware, at $2,266,667 for “developing software tools, products and analysis focused on global security.” Irregular, the offensive-cyber evaluator, is “formerly Pattern Labs,” and the UK procurement record names “Pattern Labs Tech Inc.” at the same Wilmington address as its supplier (Irregular; ProPublica, Good Ventures FY2024 Form 990-PF).
• The founder’s seat at a lab, and the director’s firm. Moskovitz’s name is on Anthropic’s May 2021 Series A announcement (“The round included participation from James McClave, Dustin Moskovitz”), and in October 2025 he said “I’m a board observer at Anthropic and we are early donors to OpenAI” (Anthropic; Stratechery). The foundation’s director Divesh Makan is named in Anthropic’s 2 September 2025 release as “Partner at ICONIQ,” which led that company’s $13 billion Series F (“Anthropic has completed a Series F fundraising of $13 billion led by ICONIQ”) (Anthropic, Series F). ICONIQ is Makan’s firm, not the foundation’s, and no page cited links the foundation’s grants to either investment. See Anthropic and OpenAI.
• Asana. Asana’s 2026 proxy statement lists “1,720,916 shares of Class A Common Stock held of record by Good Ventures Foundation” among the holdings attributed to Moskovitz, and says “Mr. Moskovitz and his spouse serve as directors on the board of Good Ventures Foundation and may be deemed to have shared voting and dispositive power with respect to the shares held by the Good Ventures Foundation” (Asana, 2026 Proxy Statement).
• A name that is not this one. AVERI’s funder list includes “Good Forever Foundation” (AVERI, about). No page cited ties it to Good Ventures Foundation and none is asserted. See AVERI.

Actions & Leadership Choices

CONDUCT: STRUCTURAL — ONE DONOR’S ENDOWMENT PAYING THE REFEREES DIRECTLY AND THROUGH ITS ADVISER. No wrongdoing is alleged. The structural point is routing: one person’s trust funds a foundation, the foundation pays some of the evaluators and pipeline organizations itself and pays the rest on an advisor’s recommendation, and two of the foundation’s three directors sit on the board of the advisor, which also advises other donors into the same fields. The foundation’s own best case is that it has said all of this on its pages: it has no staff, it uses an advisor, it names its board, it prints every grant, and its advisor tells the world that Good Ventures is “a concentrated share of AI safety philanthropic funding” and asks others to join.

The inflow is larger than the outflow. In the year to 30 June 2025 contributions of $1,395,845,381 met grants of $306,706,819, and total assets rose from $5.86 billion (June 2023) to $10.11 billion. The founders’ letter and the advisor describe a commitment to spend down; the return shows an endowment that grew over the period. Both are true on their own pages. The return also shows the foundation paying out more than the payout rule required in earlier years: Part XII carries $317,576,940 of excess distributions forward from 2021, 2022 and 2023.

The foundation pays names the advisor’s listings omit, and the reverse. Two payments on the return equal grants in the advisor’s records to the dollar ($2,321,550 and $863,143); the MATS lines appear on the return and on MATS’s page under “Good Ventures Foundation” while the advisor’s search listing carries MATS grants of $3,382,029 and $660,000. A reader following the advisor’s grant database alone would see part of what the foundation paid; a reader following the return alone would not see what the advisor lists for other donors.

What is not on the return. No line names METR, Apollo Research, Redwood Research, Palisade Research or Transluce. The advisor’s own grant records list Redwood Research ($10,700,000 over 18 months; $5,300,000; $1,100,000), Apollo Research ($1,535,480 and $2,178,700) and Palisade Research ($1,680,000) as grantees (Coefficient Giving, Redwood; Coefficient Giving, Apollo; Coefficient Giving, Palisade), so those grants were paid either from another donor’s account, from a year outside this return, or through an intermediary the return does not name; no page cited says which.

The scope limit, stated by the advisor. Its post says some conservative groups working on AI policy “may not be a good fit for Good Ventures” and that policy groups gain from “identifiable funding sources that fit their local and political context.” That is the advisor describing a boundary on the foundation’s funding, in its own words (Coefficient Giving, Oct 2 2025).

Alliance Map

These run at the same time. One donor’s trust funds one foundation; the foundation pays the evaluators SecureBio, RAND and the legal entity of Irregular, and the pipelines MATS and ARENA; its advisor, governed by two of its three directors, recommends the grants and advises other donors into the same funds; the donor is a named funder of the fund that also pays SecureBio; and the donor, his foundation and a director’s firm each hold a documented position at the lab whose models those evaluators test. The foundation, the advisor, MATS, SecureBio, RAND and Anthropic disclose the individual ties themselves. The overlap is the finding.

Reach Assessment

Institutional: $306,706,819 in grants paid and $87,998,087 more approved in one tax year; $10,107,955,038 in assets; payer on $55,341,516 of AI-named lines to organizations that include two AI Evaluator Forum members (RAND, SecureBio), the talent pipelines that feed the field and the legal entity of a frontier-security evaluator; main funder of an advisor that says its AI commitments were $351 million in 2025. Memetic: The foundation’s own pages list a mission sentence and a letter; its worldview reaches the public through the advisor’s essays and through the press pages it links (Forbes, Vox, Fortune, the Financial Times). Its home page links no social account, and its press address is [email protected] (Good Ventures; Good Ventures, In the News).

Civilizational: For scale, the advisor’s own estimate is that philanthropic funding for climate risk mitigation was “~20x as much as philanthropic funding for AI safety and security in 2024,” and it says Good Ventures and itself are “a concentrated share” of the AI line; the foundation’s AI-named lines of $55.3 million in one year sit beside the advisor’s $351 million of 2025 commitments across all donors (Coefficient Giving, Oct 2 2025; Emily Oehlsen).

Score note. Attack 7, defense 9 and hp 8 on first-hand receipts: the Form 990-PF’s grants, Schedule B contributor and balance sheet, the foundation’s own letter and governance page, and the advisor’s own statements. Peers: Open Philanthropy (Coefficient Giving) (8/8/9, hazard 80) is the operation, with 150 staff, other donors’ money and multi-year commitments to the same grantees, so reach is one point higher and entrenchment one point higher than the foundation; Survival and Flourishing Fund (6/8/6, hazard 67) allocates a $63.13MM round for two named donors and is one point lower than this foundation’s 7/9/8 on reach and insulation and two lower on entrenchment; Schmidt Sciences (6/8/7, hazard 70) is the other family funder and prints a $1.44 billion balance sheet against this one’s $10.1 billion, one point lower on each axis. At person level, Dustin Moskovitz (8/10/9, hazard 88) scores above the foundation on reach, insulation and durability because his holdings and seats outlast and exceed any one foundation’s return. Hazard 80 sits level with Coefficient Giving: reach and entrenchment one lower, insulation one higher.

Score basis. Atk 7 is reach: the return prints $306,706,819 in grants paid, 107 lines with AI, LLM, machine-learning or alignment purposes totalling $55,341,516, and named lines to MATS Research ($4,558,450), RAND ($5,166,667), the Centre for the Governance of AI ($2,995,207), SecureBio ($1,630,000) and Pattern Labs Tech Inc ($2,266,667), and its advisor calls it “our largest funding partner” for an AI line of $351 million in 2025; it is 7 and not 8 because most lines are other causes and no statute or procurement rule makes any grantee’s output binding (ProPublica, Good Ventures FY2024 Form 990-PF; Coefficient Giving, Oct 2 2025; Emily Oehlsen). Def 9 is funding insulation: Schedule B names one contributor, the Dustin A Moskovitz Remainder Interest Trust, at $1,395,695,354 of $1,395,845,381, the four officers are paid $0, and no lab, agency or customer appears as a payer; it is not 10 because one donor supplies the money and holds an Anthropic interest and the Asana stake in which the foundation holds shares (ProPublica, Schedule B; Asana, 2026 Proxy Statement). Hp 8 is entrenchment: $10,107,955,038 in assets, “15+ years of giving,” $87,998,087 approved for future payment and an advisor that treats it as its principal partner would each take years to run off, against founders who say they will move “the last dollar out the door,” no foundation staff and no statute holding it in place (Good Ventures, Our Story; Good Ventures, Our Giving). Peers on the same values: Coefficient Giving at 8/8/9, hazard 80, holds the hazard with reach and entrenchment one higher and insulation one lower; Schmidt Sciences at 6/8/7 and SFF at 6/8/6 hold lower values on reach and entrenchment. The values hold.

Sources: Good Ventures; Good Ventures, Governance; Good Ventures, Our Story; Good Ventures, Our Giving; Good Ventures, In the News; ProPublica, Good Ventures Foundation; ProPublica, Good Ventures FY2024 Form 990-PF; ProPublica, Schedule B; Coefficient Giving, Open Philanthropy is now Coefficient Giving; Coefficient Giving, Dustin Moskovitz; Coefficient Giving, AI safety and security need more funders; Coefficient Giving, Governance; Coefficient Giving, Our Progress in 2024; Coefficient Giving, Navigating Transformative AI; Coefficient Giving, We’re urgently scaling our work on AI and biosecurity; Coefficient Giving, MATS grants; Coefficient Giving, Princeton grants; Coefficient Giving, RAND grants; Coefficient Giving, Redwood grants; Coefficient Giving, Apollo grants; Coefficient Giving, Palisade grants; Coefficient Giving, SecureBio grants; Coefficient Giving, About; MATS, Transparency; Epoch AI, Transparency; SecureBio, GPT-6 Astra assessment; Sentinel Bio, SecureBio grant; Sentinel Bio, RAND grant; SFF-2026; Anthropic, Series F; Anthropic, Series A; Stratechery, Moskovitz interview; Asana, 2026 Proxy Statement; AVERI, about; The Giving Pledge