Defining global catastrophic risk
Beacon uses a straightforward definition: global catastrophic risk is anything that could kill or cause the massive suffering of a billion or more people in a single event or tightly coupled chain of events.
This includes both existential risks (threats to humanity's long-term survival or potential) and catastrophic risks that fall short of extinction but would cause suffering at a civilisational scale. It includes risks from suffering itself (s-risks), not only risks from death.
This definition is deliberately concrete. It guides every project evaluation: does this work plausibly reduce the probability or severity of a billion-scale catastrophe? If yes, it's in scope. If the connection is too indirect or speculative, we'll help you find a better home.
Risk categories Beacon prioritises
We allocate our attention and resources based on where we believe the marginal impact is greatest, weighing the importance of the risk, the tractability of interventions, and how neglected the area is relative to its stakes.
AI safety and AI governance
This is where we expect the large majority of our portfolio to sit. The intensity and near-term nature of AI risk, combined with the rapid scaling of capabilities, makes this the area where good projects face the most acute infrastructure needs.
Beacon's scope within AI safety is deliberately broad. We support work across the full spectrum: technical alignment research, interpretability, evaluations, governance and policy, and field-building. We also welcome foundational and theoretical work that may not produce near-term legible outputs but represents serious attempts to understand the underlying problem. Even short timelines tend to have wide distributions, and we do not want to constrain ourselves into adopting particular stances on timing and usefulness of research plans in a field of high epistemic variance and uncertainty.
High-variance research agendas
Some of the most important research agendas in this space are fragile: they benefit from patient, low-pressure environments where researchers can pursue difficult questions without being optimised toward funder-legible milestones. Beacon is designed to be a home for that kind of work, and we are lucky there are funders in this space who recognise the need for these kinds of projects.
We aim to take a hits-based approach to project selection. We would rather sponsor ten serious attempts at hard problems, several of which produce negative results or pivot, than restrict ourselves to projects with predictable outputs. The field's track record suggests that breakthrough insight is more likely to come from researchers who are allowed to directly engage with the most difficult aspects of the problem than from those channelled toward incrementally publishable results.
Our scope extends beyond existential risk narrowly defined. We are concerned with catastrophic outcomes at billion-plus scale, which includes scenarios involving mass suffering, destruction of critical infrastructure, disruption of economic and institutional systems, and the coordination failures that could arise as artificial agents assume greater roles in economies and governance. Research agendas addressing the coordination of agents with varying degrees of agency, moral patienthood, and institutional authority are within scope, as is work on preventing large-scale suffering that falls short of extinction.
Biosecurity and pandemic preparedness
Engineered pandemics, dual-use biotechnology, and the growing intersection of AI capabilities with biological risk represent catastrophic threats. Actively funded research areas include pathogen detection and surveillance, gene synthesis screening, medical countermeasure platforms, respiratory protection technologies, and governance of dual-use research. There are currently no biosecurity-specific fiscal sponsors serving researchers working on these problems.
Beacon does not currently have deep in-house biosecurity expertise, and our board's primary domain knowledge is in AI safety. We are equipped to support biosecurity projects that come to us through grantmakers with strong technical taste and established due diligence processes. We expect there to be some natural domain cross-expertise in some areas, for example, the governance and management of infohazards. Over time, we intend to build domain competence through exposure to well-vetted projects, relationships with biosecurity-focused funders, and ongoing engagement with the literature. We consider this a gap we are actively working to close.
Nuclear risk
We acknowledge nuclear risk as a genuine catastrophic threat, but it is substantially less neglected than AI safety or biosecurity. There are established institutions, treaties, and funding streams dedicated to nuclear risk reduction. Beacon is unlikely to be the best home for nuclear-focused work unless it involves clear intersections with AI (e.g., AI-enabled command and control risks, autonomous weapons systems) or other emerging technology risks.
Emerging technology risks and civilisational resilience
Catastrophic risk doesn't stand still. We're open to sponsoring work on emerging threats that don't fit neatly into existing categories: novel dual-use technologies, engineered risks at scale, or civilisational resilience work that directly addresses recovery from and robustness to catastrophic events. These are evaluated case by case.
Longtermism-adjacent and cross-cutting work
Some valuable work doesn't map cleanly to a single risk category but contributes to our understanding of or ability to respond to catastrophic risk. Field-building, forecasting, institutional design for existential risk governance, and similar cross-cutting efforts may be in scope depending on the specifics of the project. We evaluate these on a case-by-case basis with a higher bar for demonstrating GCR relevance.
What falls within scope
In general, Beacon sponsors work that:
- Has a plausible, articulable connection to reducing the probability or severity of a billion-scale catastrophe
- Requires a nonprofit structure (fiscal sponsorship, 501(c)(3) status, grant administration)
- Is led by people with the skills and track record to execute on the proposed work
- Would benefit from Beacon's GCR-specific domain expertise, network, and infrastructure
This includes research, policy, field-building, technical development, evaluations, community infrastructure, and rapid-response efforts, provided they meet the GCR relevance bar.
What falls outside scope
The following are important areas of work that Beacon is not well-placed to support. We respect these efforts and will help you find sponsors better suited to them.
- Climate change (in isolation). Climate work is critical, but it operates on different timescales and has a mature ecosystem of dedicated funders and organisations. Beacon's infrastructure doesn't add value here. The exception would be climate scenarios that interact with other catastrophic risks at the billion-scale threshold.
- Local disaster response. Emergency response and disaster relief require specialised operational infrastructure that Beacon doesn't have. We can offer high-level guidance on scaling coordination across response efforts, but direct disaster response is outside our wheelhouse.
- General poverty reduction and development. Valuable work, but too broad for Beacon's focused model. Organisations like GiveWell and GiveDirectly have deep expertise here.
- Work without a clear GCR connection. If the link between your project and billion-scale catastrophic risk requires more than a few sentences to explain, we're probably not the right fit.
Edge cases
Some of the most important work happens at boundaries. Beacon pays particular attention to:
- Situations that are scaling into catastrophic territory: emerging threats that aren't yet billion-scale but are on trajectories that could get there
- Work on what authorities, institutions, or treaties can bind global responses to catastrophic events
- Research into when and how previously out-of-bounds interventions become justified as threat levels escalate
- Cross-domain projects that don't have obvious homes elsewhere (e.g., AI-enabled biosecurity monitoring, governance work spanning multiple risk domains)
If you're not sure whether your project is in scope, get in touch. We'd rather have the conversation than have you self-select out of a good fit.
Alternative fiscal sponsors
If your work doesn't align with Beacon's GCR focus, or if another sponsor is a better fit for your specific needs, we would rather point you to the right home than try to make a poor match work. The following organisations accept projects in adjacent spaces.
GCR and EA-adjacent fiscal sponsors
- Rethink Priorities Special Projects. AI governance and safety, animal welfare. Comprehensive fiscal sponsorship with operational support. Currently accepting new projects.
- Ashgro. AI safety research specifically. Model A fiscal sponsorship with full financial and HR administration.
- SparkWell (AntiEntropy). Early-stage high-impact projects. Designed as a temporary home (6–24 months) for projects testing ideas before scaling or incorporating independently.
- Manifund. Fast small grants via regranting. Best for individual researchers or small projects needing rapid funding ($5–50K), not a long-term organisational home.
Generalist and broader-scope sponsors
- Social and Environmental Entrepreneurs (SEE). Broad scope. Also hosts Survival and Flourishing Projects. 6.5% fee.
- Social Good Fund. Social change, public interest. 6–8% fee. High application volume; capacity-constrained.
- The National Network of Fiscal Sponsors maintains a directory across a range of mission areas.
Note: Effective Ventures (US and UK) is winding down its fiscal sponsorship programme and is not accepting new projects. BERI provides fiscal sponsorship only for university-affiliated existential risk research groups.
If you're unsure where to go, get in touch. We're happy to help you find the right fit even if it isn't us.