Schmidt Sciences Puts $1M Behind Multi-Agent AI Safety — and Signals Where Private Research Philanthropy Is Moving as Federal Science Contracts

There is a specific kind of grant news that never trends, because the dollar figure is small next to the federal megaprograms — a $1 million research award from a private science foundation reads as a rounding error beside a $5 billion national mission. But for a working researcher, the small private award is often the more useful instrument, and the timing of one particular opportunity makes it worth a close look.

Schmidt Sciences has opened a program called "Scaling AI Safety for a Multi-Agent World," offering awards of up to $1 million, with a deadline of August 8, 2026 and applications submitted through the Schmidt Sciences portal. On its face it is a single, modestly sized call for proposals on a narrow technical topic. But read in context — against a year of federal research contraction and a proposed overhaul of the rules governing federal grants — it is a useful lens on a larger shift: private science philanthropy is stepping into gaps that federal agencies are leaving open, and researchers who understand how these funders operate have an advantage.

This piece uses the Schmidt Sciences call as a concrete anchor to explain how private research philanthropy differs from federal funding, why it is rising in relative importance right now, and how to approach it well.

What the program actually funds

The topic is narrow and current: AI safety in multi-agent environments. The framing matters. Most public AI-safety research and funding to date has concentrated on the behavior of single models — alignment, interpretability, and control of one system. But the near-future reality is systems of many AI agents interacting: negotiating, coordinating, competing, and occasionally failing in ways that no single-agent analysis predicts. Emergent behavior in multi-agent systems — collusion, cascading failures, adversarial dynamics between agents — is a genuine open research frontier, and it is exactly the kind of problem that a nimble private funder can move on faster than a federal agency can build a program around.

An award of up to $1 million is enough to fund a serious multi-year line of work for a small academic group or to seed an ambitious pilot that later attracts larger federal or industry support. That "seed-then-scale" role is one of the defining functions of private science philanthropy, and it is worth understanding structurally rather than treating each call as a one-off.

Why private philanthropy matters more in 2026

For most of the postwar era, the mental model of American research funding put federal agencies at the center and private foundations at the margins — helpful, but supplementary. That model is under strain. NIH's dramatic reduction in new awards, the proposed embedding of shifting policy priorities into federal grant terms, and the general uncertainty around federal science budgets have all made federal funding less predictable than it has been in a generation.

When the largest funder becomes less predictable, the relative value of every alternative funder rises — not because private foundations can replace federal dollars in aggregate (they cannot; the scale gap is enormous) but because they can do things federal agencies increasingly cannot:

• Move fast. A private foundation can identify an emerging problem — multi-agent AI safety, say — and open a call in months, without the multi-year process of building a federal program.
• Fund the unfashionable or the too-new. Foundations answer to their own boards and missions, not to appropriations politics, so they can back topics that have not yet earned a federal line item.
• Tolerate risk. Private science philanthropy often explicitly seeks high-risk, high-reward work that federal peer review, optimized for reliability, tends to filter out.
• Bridge gaps. A foundation grant can keep a lab and its people intact through a fallow federal-funding period — the difference between a research program that survives a lean year and one that disbands.

None of this makes private funding a substitute for a diversified federal portfolio. It makes it an essential complement — and in a year when federal timelines are wobbling, the complement is doing more of the load-bearing.

How private funders differ from federal agencies — and why it changes your approach

Researchers fluent in federal grant-writing often stumble with private funders because they carry over habits that do not translate. The two worlds reward different things.

Federal proposals are won on methodological rigor, specific aims, and fit to a published solicitation. The review is peer-driven, criteria are explicit, and the document is long, structured, and technical. The reviewer is a fellow scientist scoring against a rubric.

Private foundation proposals are won on mission alignment and vision. The foundation has a worldview — Schmidt Sciences exists to accelerate scientific discovery and, in this program, to make multi-agent AI safer — and it funds researchers who advance that worldview. The document is often shorter, the review is program-officer-driven rather than panel-driven, and the winning proposal reads less like a methods section and more like a compelling argument that this work, by this team, will move the thing the foundation cares about.

Concretely, that changes how you should approach the Schmidt Sciences call, and calls like it:

• Lead with the vision, not the methods. Open with why multi-agent AI safety is a consequential problem and why your approach is positioned to make real progress. Save the technical depth for where it earns its place.
• Demonstrate mission fit explicitly. Study the funder's stated goals and reflect them back in your framing. Program officers fund researchers who clearly share their mission.
• Show trajectory, not just a project. Foundations frequently fund people and programs as much as individual studies. Make clear where a $1 million seed could lead — the larger program it unlocks, the follow-on federal or industry funding it positions you for.
• Respect the format. Private applications are often shorter than federal ones. Concision is a feature; a padded proposal signals you did not understand the funder.

The broader Schmidt Sciences and private-funder landscape

The multi-agent AI-safety call is one of several active private-philanthropy opportunities worth knowing, and the surrounding landscape illustrates the range. Schmidt Sciences separately runs a Virtual Institute for the Carbon Cycle at the $10 million scale, with expressions of interest due in late August — evidence that these funders operate across a wide budget spectrum, from single-lab seeds to institute-scale commitments. Elsewhere, the Harry Frank Guggenheim Foundation funds distinguished scholars in the $15,000–$75,000 range, and disease-specific funders like the Robert L. Fine Cancer Research Foundation put $250,000 behind targeted pancreatic-cancer work. Each has its own mission, its own review culture, and its own application rhythm.

The lesson is not to chase any single one of these. It is to recognize that a portfolio of private funders — matched carefully to a lab's actual mission and stage — is now a core part of a resilient research-funding strategy, not a fallback. The researcher who knows which three foundations align with her work, and who has built relationships with their program officers before a call opens, is positioned in a way that no last-minute federal-only strategy can match.

What to do now

If your work touches AI safety, multi-agent systems, or the reliability of interacting autonomous agents, the Schmidt Sciences call is a concrete, near-term action item with an August 8, 2026 deadline. But the more durable move is strategic:

• If the topic fits, apply — and lead with vision. Frame the multi-agent safety problem, then your approach, then your team's fit. Keep it concise.
• Build a private-funder map for your lab. Identify the three to five foundations whose missions genuinely align with your research, across a range of award sizes.
• Engage program officers early. Private funding is relationship-driven in a way federal funding is not. A conversation before a call opens is worth more than a polished application after.
• Treat private philanthropy as core, not backup. In a year of federal uncertainty, the labs that thrive will be the ones that diversified deliberately rather than reactively.
• Watch for follow-on scale. A $1 million seed that produces results is the strongest possible position from which to win a larger federal award later — the two are complementary, not competing.

The Schmidt Sciences multi-agent AI-safety program will fund a handful of teams. But the pattern it represents — fast, mission-driven private capital moving into a frontier before federal programs arrive — is one every research organization should be built to catch. The dollar figures are smaller. The strategic value, in 2026, is not.

Granted tracks private foundation and federal research funding side by side, matched to your organization's mission and stage. Explore research funding opportunities across both worlds in one place.