Where AI Safety
Funding Goes
Tracking capital flows into AI safety research, governance, and tooling — and how they compare to the hundreds of billions invested in building AI capabilities.
$320B+ on building AI. $2B on making it safe.
In 2025, the four largest tech companies announced over $320B in AI infrastructure spending. Total tracked safety funding across philanthropy, government, and commercial investment is roughly 199x smaller.
Where Safety Funding Goes
Allocation across six categories derived from how major funders (Coefficient Giving, SFF, government AISIs, VCs) classify their own grants. Based on 29 tracked records from public grants databases, government announcements, and Crunchbase.
Top Funders
Largest sources of AI safety funding by total amount tracked. Data from public grant databases, government budget announcements, and press releases.
Largest Grants & Investments
Individual funding records by size. Includes grants, government allocations, and VC rounds.
What these numbers don't show
This database tracks publicly identifiable funding. It undercounts in every category — internal safety teams at large labs, classified government programs, and unreported philanthropic grants are all excluded. The true total is higher, though likely still orders of magnitude below capabilities investment.
Spending ratios use announced 2025 capital expenditure from the four largest AI investors (Microsoft, Alphabet, Amazon, Meta). Not all capex is pure AI — but AI is the stated driver in each case.
Allocation percentages describe where tracked funding goes, not where it should go. Relative underfunding of a category does not necessarily indicate a gap — some areas require less capital than others.
Data sources: Coefficient Giving grants database, Survival and Flourishing Fund announcements, UK/US AI Safety Institute budgets, NSF ReDDDoT, EU Horizon Europe, Schmidt Sciences, Frontier Model Forum, Crunchbase, corporate earnings calls.
Mappera field assessment v0.3 — funding data updated as records are identified.