SCHMIDT SCIENCES

Institutional Archetype

THE FAMILY FUNDER OF THE FELLOWS — Schmidt Sciences is the science philanthropy of Eric and Wendy Schmidt. In AI it runs three instruments: a safety-science grant program, the AI2050 fellowships, and a seat on the pooled funds that the UK government’s AI Security Institute and the labs’ own Frontier Model Forum run. Its name sits on the funder lists of the evaluators the labs cite: METR, Epoch AI, Stanford CRFM and FAR.AI’s chief executive. The founder, Eric Schmidt, is also a personal investor in a lab and a founding partner of an industry safety-tooling body. The file is open because the same family money sits on the funder side of the evaluators and beside the labs on the pooled funds. This profile scores the position, not the family.

Mandate & Origin

The organization’s own text: “Schmidt Sciences is part of the philanthropic organizations and initiatives created and funded by Eric and Wendy Schmidt,” and “We envision a world where science is bold, collaborative and provides well-researched solutions for planetary challenges” (Schmidt Sciences, About). Its FAQ gives the form: “Schmidt Sciences, LLC is a 501(c)3 non-profit organization supported by Eric and Wendy Schmidt,” and “Founded in 2024, Schmidt Sciences brings together the science-focused efforts of Eric and Wendy Schmidt” (Schmidt Sciences, FAQ). The same page says it is “a subsidiary of The Eric and Wendy Schmidt Fund for Strategic Innovation (EIN: 46-3460261).” Its work runs through five centers: AI & Advanced Computing, Astrophysics & Space, Biosciences, Climate, and Science Systems (Schmidt Sciences).

Eric Schmidt’s own biography names the same founding and lists the earlier vehicles separately: “Eric and his wife Wendy have founded several philanthropic organizations,” including “the Schmidt Family Foundation, Schmidt Ocean Institute, and Schmidt Futures,” and “Most recently in 2024, they founded Schmidt Sciences” (ericschmidt.com). The address schmidtfutures.org now redirects to schmidtsciences.org (schmidtfutures.org). Stanford’s Center for Research on Foundation Models still thanks “Schmidt Futures” by that name among its supporters (Stanford CRFM, Support), so the older name is still the one some recipients publish.

Leadership. The team page lists Eric and Wendy Schmidt as “Founders,” Stuart Feldman as “President” and Mark Greaves as “Vice President” (Schmidt Sciences, Team). Its AI safety program announcement calls Feldman “chief scientist and president” and names Michael Belinsky as the program’s director and lead (Schmidt Sciences, Feb 11 2025). The parent foundation’s FY2024 filing, as ProPublica lists it, names Eric Schmidt as President and Director and Wendy Schmidt as Director with no compensation, and Feldman, Greaves, Michael Gable and Charles Giller among officers (ProPublica Nonprofit Explorer, EIN 46-3460261).

Funding & Backers

The money is the founders’. ProPublica’s summary of the parent foundation’s Form 990-PF filings, which run every year from 2013 to 2024, shows for the year 2024 revenue of $406,766,399, expenses of $313,041,582 and net assets of $1,440,351,492, with charitable disbursements of $315,860,401 in 2024 and $254,680,317 in 2023 (ProPublica Nonprofit Explorer, EIN 46-3460261). The foundation’s tax exemption dates from September 2014. The figures cover the whole parent, not AI alone; the AI share is not separately published on the pages read.

Schmidt Sciences says “We proactively seek out grant opportunities and do not accept unsolicited proposals” (Schmidt Sciences, FAQ). Its AI2050 page describes “Eric and Wendy Schmidt’s $125 million commitment over five years” and says “Fellows are eligible to receive up to $300,000 over two years” (Schmidt Sciences, AI2050).

Who it pays, as the recipients record it.

Institutional Voice & Intent

The organization’s register is the research agenda and the press release, and its subject is that safety science is under-supplied.

• The trustworthy-AI page: “The science of trustworthy AI is nascent.” It says “commercial incentives for foundational, pre-product safety science are often weaker than incentives for capability and product improvements,” that “safety research for the largest AI models is primarily conducted by leading AI labs,” and that foundational advances “function as a global public good, motivating targeted philanthropic support for academic and nonprofit researchers” (Schmidt Sciences, Science of Trustworthy AI).
• The 2026 agenda: “This agenda outlines Schmidt Sciences’ 2026 priorities for technical research on trustworthy AI,” and one of its three aims is to “Oversee AI systems with superhuman capabilities and address multi-agent risks” (Schmidt Sciences, Research Agenda).
• Eric Schmidt, in the February 2025 announcement: “With efforts like the AI Safety Science program, we can help build a future in which AI benefits us all.” Feldman: “This is more than a funding program.” Belinsky: “The science of AI safety is a crucial new field that is underfunded” (Schmidt Sciences, Feb 11 2025).

Stated intent: Fund the foundational science of AI safety that the market and the labs under-supply, and support the researchers who do it.

Observed intent: Seed a research community through named programs and fellowships, join the pooled funds beside the labs and a government institute, and supply the evaluation layer with money, while the founder separately invests in a lab.

Gap: The organization’s own page says safety research “is primarily conducted by leading AI labs” and that philanthropy fills the gap; its funder lists show the labs on the pooled funds it joins and OpenAI supplying the API access for its own program. What the record does not show is any published mechanism by which a grantee’s findings are insulated from the funder, and none is claimed. The gap is structural, not a finding about any grantee.

Interest: Schmidt Sciences does not describe its AI positions as self-interest and the record does not say they are. What each position would pay the funder or its principals is set out from its own pages and the filings; no fee or return to the foundation is documented.

• Safety science as a philanthropic public good. Position, the organization’s own: foundational advances “function as a global public good, motivating targeted philanthropic support for academic and nonprofit researchers” (Schmidt Sciences, Science of Trustworthy AI). What it pays: the program picks its own agenda and advisory board (the page lists Percy Liang, Yonadav Shavit, Ajeya Cotra, JueYan Zhang and Mark Greaves), and the grantee community it funds supplies the evaluations the labs cite; its founder separately appears among participants in Anthropic’s 2021 Series A (Anthropic), a documented tie that sits beside the position and is not claimed to cause it. Best case, the organization’s: “commercial incentives for foundational, pre-product safety science are often weaker than incentives for capability and product improvements.” Convergence: the Alignment Project pools Schmidt Sciences with OpenAI, Microsoft, AWS and Anthropic, and the AI Safety Fund with the labs, so the labs and the philanthropy fund the same pots (Alignment Project; Frontier Model Forum).
• Measurement science for frontier systems. Position: Belinsky, “The science of AI safety is a crucial new field that is underfunded” (Schmidt Sciences, Feb 11 2025). What it pays: METR, Epoch AI and the AI2050 and Trustworthy AI awardees are recipients of record, so a field organized around evaluation is a field in which its grantees are the instruments (METR; Epoch AI). Best case, the organization’s: the field is “underfunded” and the grantees need a source that is not a lab. Convergence: Coefficient Giving also funds the same community and sits beside Schmidt Sciences at Ohio State’s faculty briefing; see Open Philanthropy and The AI Evaluators.

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 other checkbook. Coefficient Giving’s AI commitments are reported in hundreds of millions a year (Open Philanthropy); Schmidt Sciences’ published AI programs are a $10 million safety science cohort, a 2026 request of “up to $5M+”, and $125 million over five years for fellowships across AI2050 (Schmidt Sciences, AI2050). The two sit on the same pages: Schmidt Sciences’ trustworthy-AI page lists Ajeya Cotra as “Senior Program Officer at Open Philanthropy” and JueYan Zhang as “Fund Manager, AI Safety Tactical Opportunities Fund” (Schmidt Sciences, Science of Trustworthy AI), and the Alignment Project lists the AI Safety Tactical Opportunities Fund among its funders.
• A seat on the government’s fund. The UK AI Security Institute’s Alignment Project is “a global fund of over £27 million” “Backed by an international coalition of governments, industry, venture capital and philanthropic funders,” and it names Schmidt Sciences beside OpenAI, Microsoft, Amazon Web Services and Anthropic (Alignment Project). See UK AI Security Institute.
• A seat on the labs’ fund. The Frontier Model Forum’s AI Safety Fund names Schmidt Sciences, and Jaan Tallinn, among its philanthropic supporters (Frontier Model Forum). See Frontier Model Forum and Jaan Tallinn.
• The evaluators. METR lists it among its foundation donors; Epoch lists it for a December 2025 pilot; CRFM thanks Schmidt Futures. METR’s page says it “has not accepted funding from AI companies” and lists the UK AI Security Institute and Packard among its donors (METR). See METR, Epoch AI and Stanford CRFM.
• The founder’s other roles. Eric Schmidt’s bio lists him as “Chair and CEO of Relativity Space” and says “he launched, and currently Chairs, the Special Competitive Studies Project” (ericschmidt.com); his personal tie to ROOST is “ROOST’s founding partners include Eric Schmidt, Discord, OpenAI, Google, Roblox” (PR Newswire, Feb 10 2025); the foundation is not named as a ROOST partner in that release. AI2050 is “an initiative of Schmidt Sciences, conceived and co-chaired by Eric Schmidt and James Manyika” (Schmidt Sciences, AI2050). The full map of the founder’s seats is in Eric Schmidt and ROOST.

Actions & Leadership Choices

CONDUCT: STRUCTURAL — FUNDS THE EVALUATORS, AND THE FOUNDER FUNDS LABS. No wrongdoing is alleged. The structural point is placement: one family’s money is on the donor list of the outside evaluators, in the pooled fund the UK government runs, in the pooled fund the labs run, and in the fellowships of several researchers who study the labs; the founder separately invests in a lab and sits on an industry safety-tooling body. The organization lists its recipients itself, which is the strongest case for the arrangement.

The earlier vehicle, in the news. In March 2022 Politico reported that Schmidt Futures had indirectly paid the salaries of two White House Office of Science and Technology Policy staffers; CNBC reports that Eric Schmidt called that claim “largely false” (CNBC, Mar 29 2022). The report concerned Schmidt Futures, not Schmidt Sciences, and is recorded here as the reporting and the denial.

Alliance Map

These run at the same time. One family fund pays the fellows who study the labs, sits on the evaluators’ donor lists, joins the pooled fund the UK government runs and the pooled fund the labs run, takes compute and API access from a lab and a safety institute, and shares an advisory list with another philanthropy’s staff, while its founder invests in a lab and sits on an industry safety-tooling body. The organization lists most of those ties itself or through the recipients. The overlap is the finding.

Reach Assessment

Institutional: A $10 million safety-science cohort, a 2026 request of up to $5M+, a $125 million fellowship commitment, and named seats on two pooled funds. Memetic: The framing in its own pages, that safety science is “nascent” and under-supplied by markets, is the framing a grantee community repeats; the page does not say who adopts it.

Civilizational: Not separately measurable from the pages read; the parent foundation’s $1.44 billion in net assets and $315,860,401 in 2024 disbursements bound what the family can direct across all five centers, of which AI is one (ProPublica).

Score note. Attack is 6, defense 8 and hp 7 on first-hand receipts: the February 2025 program announcement, the Ohio State recap of the 2026 request, the funder lists of METR, Epoch, the Alignment Project and the AI Safety Fund, and the parent’s Form 990-PF figures. Hazard 70. Peers: Open Philanthropy (8/8/9, hazard 80) is the one-rung-up funder with a larger published AI commitment and more multi-year grants; Gates Foundation (6/9/9, hazard 80) holds the same reach with a larger endowment and more insulation; Redwood Research (6/8/7, hazard 66) holds the same values at a lower hazard because it is a grantee, not a funder of others. Hazard 70 sits below Open Philanthropy because its published AI money is far smaller.

Score basis. Atk 6 is reach: the organization’s own announcement puts “$10 million” into 27 projects with compute from the Center for AI Safety and API access from OpenAI, its AI2050 page puts “Eric and Wendy Schmidt’s $125 million commitment over five years” behind fellowships, and the Alignment Project and the AI Safety Fund name it among their funders (Schmidt Sciences, Feb 11 2025; Schmidt Sciences, AI2050; Alignment Project; Frontier Model Forum). The points under 7 are that nothing obliges a lab or an agency to use any grantee’s output and that no AI-specific yearly total is published. Def 8 is funding insulation: the operating entity is “a subsidiary of The Eric and Wendy Schmidt Fund for Strategic Innovation,” whose FY2024 filing shows $406,766,399 in revenue and $1,440,351,492 in net assets and names Eric Schmidt as President and Director, so no lab or agency pays the funder (Schmidt Sciences, FAQ; ProPublica). The points under 9 are concentration in one family’s fortune and the founder’s personal Anthropic participation (Anthropic). Hp 7 is entrenchment: Form 990-PF filings from 2013 to 2024, $1.44 billion in net assets, five centers and a multi-year AI2050 commitment would each take years to run off, against the fact that no statute or procurement vehicle holds any of it in place and the founders can redirect it (ProPublica; Schmidt Sciences, AI2050). Peers: Open Philanthropy at 8/8/9, hazard 80, and Gates Foundation at 6/9/9, hazard 80. The values hold.

Sources: About — Schmidt Sciences; FAQ — Schmidt Sciences; Team — Schmidt Sciences; Schmidt Sciences; Science of Trustworthy AI — Schmidt Sciences; Trustworthy AI Research Agenda — Schmidt Sciences; New $10 Million AI Safety Science Program (Feb 11, 2025) — Schmidt Sciences; AI2050 — Schmidt Sciences; Awardees — Schmidt Sciences; ericschmidt.com, bio; schmidtfutures.org; METR, About; Stanford CRFM, Support; Alignment Project, About; Frontier Model Forum, AI Safety Fund; Epoch AI, Transparency; FAR.AI, Adam Gleave named AI2050 Early Career Fellow (Nov 4, 2025); ROOST launch release — PR Newswire (Feb 10, 2025); Anthropic, Series A (2021); CNBC (Mar 29, 2022); The Eric and Wendy Schmidt Fund for Strategic Innovation — ProPublica Nonprofit Explorer; Event recap, Schmidt Sciences and Coefficient Giving — Ohio State (Mar 24, 2026).