Mission & Theory of Change

Why Beacon exists and how it creates impact.

Problem statement

The landscape of global catastrophic risk research is growing rapidly but remains structurally underfunded and operationally underserved. Open Philanthropy committed roughly $50M to technical AI safety research alone in 2024 and described that as "too slow." Coefficient Giving estimates that AI safety philanthropy is roughly 1/20th the scale of climate philanthropy despite comparable or greater stakes. The talent pipeline is expanding: MATS trains 100+ aspiring safety researchers annually, CAIS runs compute clusters supporting hundreds of researchers, and new policy organisations are standing up globally.

But the infrastructure supporting these efforts hasn't kept pace. Researchers and project founders face a choice between:

The results are opportunity cost, preventable delays, and bottlenecks on getting good research done.

Gap analysis

Existing fiscal sponsors serve the nonprofit ecosystem well in general, but GCR projects face specific frictions that generalist infrastructure doesn't address.

Generalist fiscal sponsors process applications without domain expertise. They can't meaningfully evaluate whether a project's theory of change is credible within the GCR landscape, which means they add administrative overhead without adding strategic value. Project founders get a tax-exempt wrapper but no signal boost, no network effects, and no quality filtering that would give funders confidence.

The few GCR-adjacent fiscal sponsors that do exist tend to be capacity-constrained, US-centric in their networks, or focused on specific sub-domains (AI safety only, biosecurity only) rather than the cross-cutting nature of catastrophic risk. Projects that sit at intersections (say, AI-enabled biosecurity monitoring, or governance work spanning multiple risk domains) often don't have obvious homes.

GCR work also has characteristics that make it a poor fit for conventional academic or institutional settings. Researchers struggle to find homes for their agendas due to perceived reputational risk, while projects routinely face unusual administrative demands: coordinating international collaborations across jurisdictions, managing information hazard considerations in dual-use research, navigating relationships with government and intelligence community stakeholders, or operating under timelines driven by capability developments rather than academic calendars.

Meanwhile, the GCR funding ecosystem relies heavily on a small number of large funders (Open Philanthropy, the Survival and Flourishing Fund, Longview Philanthropy) who have expressed a desire for more independent capacity in the space. Open Phil has explicitly said that other funders "can correct our blind spots and make new bets." But new funders entering the space face their own infrastructure gap: they need credible intermediaries who can vet projects, manage funds, and provide assurance that money is being well-spent on genuine GCR work.

Beacon's intervention model

Beacon combines fiscal sponsorship with deep GCR domain focus to create value that neither generalist sponsors nor standalone projects can achieve on their own. The core mechanism works on three levels.

Quality filtering

By maintaining a high bar for GCR relevance and actively redirecting out-of-scope projects to better-suited sponsors, Beacon creates a curated portfolio. When a project operates under Beacon, that signals to funders that someone with domain expertise has evaluated its GCR relevance and operational viability. This is a credibility function that generalist sponsors can't provide.

Operational infrastructure

Beacon handles the 501(c)(3) compliance, financial administration, reporting, and governance burden so project founders can focus on their actual work, whether that's research, field-building, policy, or technical development. This is standard fiscal sponsorship, but done by people who understand the domain well enough to anticipate needs (like navigating funder reporting requirements specific to the GCR space, or understanding why a project's timeline might shift based on developments in AI capabilities).

Network concentration

By housing multiple GCR projects under one roof, Beacon creates proximity effects. Sponsored projects share a funder ecosystem, face similar governance challenges, and can learn from each other's operational patterns. Over time, Beacon's resource library, templates, and institutional knowledge about what works in GCR project management become a shared asset that no individual project could build alone.

Impact pathways

Beacon doesn't reduce catastrophic risk directly. Its impact is mediated through the projects it sponsors, and the degree to which better infrastructure enables those projects to be more effective.

Reducing time-to-impact for new projects

A researcher or organiser with a fundable GCR idea can begin operating within weeks under Beacon, rather than spending 6–12 months incorporating, filing for 501(c)(3) status, and building administrative infrastructure. In a field where timelines may be short and the window for intervention is uncertain, this acceleration matters.

Enabling projects that wouldn't otherwise exist

Some valuable GCR work is too small, too early-stage, too unconventional, or too cross-disciplinary to justify standing up a new nonprofit. Fiscal sponsorship lets these projects exist at all: pilot studies, time-limited field-building efforts, exploratory research programmes, rapid-response policy work.

Improving funder confidence and capital deployment

By providing a vetted, GCR-focused pipeline of projects with professional financial administration and reporting, Beacon makes it easier for both existing and new funders to deploy capital efficiently. This is particularly relevant for funders entering the GCR space who don't yet have their own evaluation capacity.

Building institutional knowledge

Over time, Beacon accumulates data on what kinds of GCR projects succeed, what common failure modes look like, and what operational support makes the biggest difference. This knowledge, shared through resources and guidance, raises the baseline quality of new GCR projects entering the ecosystem.

Assumptions and uncertainties

For Beacon's model to work, several things need to be true.

There must be a meaningful pipeline of GCR projects that need fiscal sponsorship. We believe this is the case based on growth in the field, the expressed need from funders for more independent capacity, and the gap in domain-specific fiscal sponsor availability.

Quality filtering must add real value. The bar for GCR relevance needs to be high enough to be meaningful but not so narrow that it excludes good work at the margins. Getting this calibration right will require iteration, and we expect our criteria to evolve as the field evolves.

The GCR funding landscape must remain healthy enough to support sponsored projects. Beacon provides infrastructure, not funding. If the broader funding environment contracts significantly, our sponsored projects would struggle regardless of the quality of our operations. We're somewhat insulated by the diversity of GCR funders, but concentration risk in the funding ecosystem is a real concern.

Beacon itself must remain operationally sustainable. Our model depends on admin fees from sponsored projects, which means we need sufficient volume and project scale to cover our own costs, supplemented by retained earnings on funds under management. We've modelled this and believe it's viable, but early-stage nonprofit economics are inherently uncertain.

Our definitions must evolve with the threat landscape. "Global catastrophic risk" will change as capabilities and threats evolve. Beacon needs to be adaptive enough to sponsor work on emerging risks that aren't yet well-defined, and to accept a reasonable rate of negative results and honest failure.

Better infrastructure must actually be impactful. This model claims that GCR impact is at least partly operational. We believe this is true, but we acknowledge that the causal chain from "smoother fiscal sponsorship" to "reduced catastrophic risk" has many subtle and diffuse links, and our contribution is one piece of a much larger puzzle. We're comfortable with that.

Risks and challenges