How each estimate is worked out
Some grant amounts aren't published, so the site estimates them. Each estimate rests on a reason with its source. 75 reasons cover $511M of estimated grants; 10 need a look (flagged first). Figures in the reasons that come from our own rows (how many grants, their total, the range) are filled in from the data, so they stay current.
| Funder | Reason | Rows | Total | Range | Years | Check |
|---|---|---|---|---|---|---|
| Foresight Institute: Funding | Foresight reports $1.5M of AI safety grants in 2024 (foresightinstitute.substack.com) across about 20 grantees without per-grant amounts: about ~$75k each, within its stated $30k–$100k range. | 23 +1 not counted | $1.7M | $30k–$100k | 2023–2024 | ⚠ rows total $1.8M, source states $1.5M; 24 rows, source states 20 |
| OpenAI Foundation | AI-risk share of the ‘more than $130 million’ AI Resilience programme (openaifoundation.org). It names four areas with no split: AI model safety (counted in full), bio-resilience and cyber-resilience (framed partly as defence against AI misuse; counted half each) and AI's impact on young people (not counted). Equal weights give 50%; the range is 25–75%. | 1 | $65M | $33M–$98M | 2026 | ⚠ rows total $65M, source states $130M |
| The Alignment Project | AISI says the first round distributed over £27M across over 60 projects (alignmentproject.aisi.gov.uk) at £50k–£1M each, and publishes no per-project amounts. Each grant is counted at the average, £450k (£27M ÷ 60), converted at Trace's ~1.31 to ~$590k; the range is £50k–£1M. | 56 | $33M | $66k–$1.3M | 2026 | ⚠ 56 rows, source states 60 |
| Science of Trustworthy AI | Schmidt Sciences launched its $10M AI Safety Science programme by selecting 27 projects (schmidtsciences.org) (Feb 2025) without per-project amounts; an even share of $10M is ~$370k each. | 26 | $9.6M | 2025 | ⚠ 26 rows, source states 27 | |
| Astralis Foundation | From the recipient's financial report; dollar amount approximate. | 1 | $428k | 2025 | ⚠ reason only in the row's note | |
| Longview Philanthropy | Longview doesn't itemise most of the giving it advises on. These amounts are Trace's estimate of what it gave out each year for AI safety, less the grants listed here individually. | 1 | $294k | 2025 | ⚠ no source link | |
| Coefficient Giving | From the recipient's transparency table (amounts in thousands of euros); Coefficient Giving's own listing not found; dollar amount approximate. | 1 | $237k | 2024 | ⚠ reason only in the row's note | |
| Survival and Flourishing Fund | Estimate of the AI-safety share of row 53b8f2a5 (same grant, left out of totals). General-support grant; share counts GCRI publications by topic, not budget. | 6 | $136k | $2k–$54k | 2019–2024 | ⚠ reason only in the row's note |
| Jaan Tallinn | Estimate of the AI-safety share of row 80d8a0da (same grant, left out of totals). General-support grant; share is GCRI's own description of its 2019 focus, not a budget split. | 13 | $81k | $250–$77k | 2013–2019 | ⚠ reason only in the row's note |
| EA Donor Lottery (CEA / EA Funds / Giving What We Can) | Estimate of the AI-safety share of row adfa7916 (same grant, left out of totals). Unrestricted support for Seth Baum's work given late 2018; share sits between GCRI's 2018 (~40%) and 2019 (~75%) AI focus. | 1 | $10k | $6k–$15k | 2018 | ⚠ reason only in the row's note |
| German government | Germany's digital ministry (BMDS, 16 Sep 2026 (bmds.bund.de)) plans about €100M for LawZero's Berlin site, subject to EU state-aid notification; ≈ US$115.4M at that day's ECB rate. One planned commitment: no payment schedule is published. | 1 | $115M | 2026 | ✓ | |
| Canadian government (LawZero commitment) | Canada's own announcement (canada.ca) says Canada plans to invest CA$150M in LawZero through the Strategic Response Fund; no period is given. At the 16 Sep 2026 rate (CA$1 = US$0.717) that is about US$108M. | 1 | $108M | 2026 | ✓ | |
| Coefficient Giving | CG's Sept 2026 post says it recommended 'more than $70 million over the next two years' to Redwood Research. $70M is that stated lower bound. source | 1 | $70M | 2026 | ✓ | |
| The Audacious Project | METR says The Audacious Project catalysed about $38M for Project Canary (metr.org) (RAND and METR), of which about $17M supports METR. RAND's share is not published, so it is taken as the remainder: $38M − $17M ≈ $21M. | 1 | $21M | 2024 | ✓ | |
| AI2050 | Schmidt Sciences announced up to $16M for the 2026 cohort (schmidtsciences.org) of 28 fellows (6 senior, 22 early-career). Trace weights senior fellows ~3.5x early-career, giving ~$1.3M per senior and ~$370k per early-career fellow (6×1.3M + 22×370k ≈ $15.9M). Since $16M is a cap ('up to'), the cohort may receive less in total, and any one fellow may get more or less than this share. | 9 +19 not counted | $6.1M | $370k–$1.3M | 2026 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (90%, range 75–100%) of a grant for software and analysis on the security risks of advanced technologies. Pattern Labs, now Irregular (irregular.com), calls itself a frontier security lab protecting the world from increasingly capable AI systems. | 1 | $6.1M | $5.1M–$6.8M | 2024 | ✓ |
| Macroscopic Ventures | Macroscopic (then CERR) committed $15M to seed the Cooperative AI Foundation (cooperativeai.com), announced May 2021. No payment schedule is published, so it is spread evenly over 2021–2025: ~$3M a year. The 2021–23 shares are counted at the Cooperative AI Foundation, as its own grants. | 2 +3 not counted | $5.4M | $2.4M–$3M | 2023–2025 | ✓ |
| Science of Trustworthy AI | Schmidt Sciences' inference-time compute RFP offered up to $500k per project (lesswrong.com) and published no round total; Trace counts each of the 21 funded projects at ~$250k, half the cap. | 21 | $5.3M | $250k–$500k | 2025 | ✓ |
| Foresight Institute: Funding | Foresight names its grantees but not amounts and says typical grants are $30k–$100k (foresight.org). Each of the 93 grants dated 2025–2026 is counted at that range's geometric midpoint, ~$55k. | 93 | $5.1M | $30k–$100k | 2025–2026 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (60%, range 40–80%): Lightcone calls itself ‘the primary infrastructure provider for the rationality and AI safety communities’ (lesswrong.com) and says ‘the single biggest topic on LessWrong is artificial intelligence’; LessWrong (lesswrong.com)’s purpose is broader (improving reasoning). No split by topic is published; the 2023 grant puts $2.59M of $2.96M into LessWrong. | 4 | $4.9M | $120k–$3.6M | 2020–2023 | ✓ |
| AI Safety Fund (Frontier Model Forum) | The Frontier Model Forum says its 11 Round 2 grantees (frontiermodelforum.org) (Dec 2025) received more than $5M through the AI Safety Fund, without per-grant amounts. An even share is ~$455k. Since $5M is a minimum, the average grant is at least that, but any one grant may be larger or smaller. | 10 +1 not counted | $4.5M | 2025 | ✓ | |
| Global Catastrophic Risks Opportunities | AI-safety share (35%, range 20–50%): Atlas names four global challenges it applies its tools to, ‘global poverty, existential risks to humanity, risks from advanced AI, or the reform of economic institutions’ (atlasfellowship.org); its 2022 programme had ‘a special focus on artificial intelligence’, and staff (atlasfellowship.org) ‘hope to empower fellows who are especially interested in this topic’ (AGI risk). No split of content or fellows by topic is published. | 3 | $3.1M | $360k–$2.5M | 2022–2023 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (70%, range 50–90%): two of BlueDot's three course areas are AI (AI safety, AI governance and biosecurity (bluedot.org); biosecurity was added in 2023). No per-course participant split is published. | 3 | $3.1M | $194k–$2.2M | 2023–2024 | ✓ |
| AI Safety Fund (Frontier Model Forum) | The AI Safety Fund says its 12 Round 1 grants ranged from $150k to $400k, totalling over $3M (aisfund.org). An even share is ~$250k; any one grant could be anywhere in that $150k–$400k range. | 12 | $3M | $150k–$400k | 2024 | ✓ |
| AI2050 | Schmidt Sciences says the 2025 cohort of 28 (schmidtsciences.org) (7 senior, 21 early-career) is eligible for more than $18M. Splitting it with senior fellows at 3.5x gives ~$1.4M per senior fellow (7×1.4M + 21×400k = $18.2M). | 2 +5 not counted | $2.8M | 2025 | ✓ | |
| Patrick J. McGovern Foundation | CSET's donor lists are cumulative: this funder moved from the $500k–$999,999 band (2019–23) to the $1M–$9,999,999 band (to 2024), so the 2024 row counts only the increase: the difference of the bands' geometric midpoints, likely $0–$9.5M. source | 1 | $2.5M | $2.5M–$9.5M | 2024 | ✓ |
| Musk Foundation | CSET's 2023 report ‘CSET at Five’ lists this funder in its $1,000,000–$6,000,000 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(1M × 6M) ≈ $2.45M. source | 1 | $2.4M | $1M–$6M | ✓ | |
| NobleReach Foundation | CSET's 2023 report ‘CSET at Five’ lists this funder in its $1,000,000–$6,000,000 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(1M × 6M) ≈ $2.45M. source | 1 | $2.4M | $1M–$6M | ✓ | |
| AI2050 | Schmidt Sciences announced up to $12M for the 2024 cohort (schmidtsciences.org) of 25 (5 senior, 20 early-career). Splitting it with senior fellows at 4x gives ~$1.2M per senior fellow (5×1.2M + 20×300k = $12M). Since $12M is a cap ('up to'), the cohort may receive less in total, and any one fellow may get more or less than this share. | 2 +3 not counted | $2.4M | 2024 | ✓ | |
| Global Catastrophic Risks Opportunities | AI-safety share (80%, range 60–95%): the grant supports a Berkeley office shared by about 20 organisations, and Constellation describes itself as enabling ‘people tackling the most important problems in AI safety & security’ (constellation.org). Its 2023 tenant list isn't published. | 1 | $2.4M | $1.8M–$2.9M | 2023 | ✓ |
| AI2050 | Schmidt Sciences says the 2025 cohort of 28 (schmidtsciences.org) (7 senior, 21 early-career) is eligible for more than $18M. Splitting it with senior fellows at 3.5x gives ~$400k per early-career fellow (7×1.4M + 21×400k = $18.2M). | 5 +16 not counted | $2M | 2025 | ✓ | |
| Global Catastrophic Risks Opportunities | AI-safety share (70%, range 50–90%): the grants pay for videos ‘about AI and AI risk, as well as other topics relating to rationality’, and the channel says it is ‘particularly focused on AI Safety and AI Alignment’ (rationalanimations.com). No count of videos by topic is published. | 3 | $1.9M | $143k–$1.3M | 2021–2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (100%, range 90–100%): the grant text (coefficientgiving.org) says Timaeus uses singular learning theory to understand how training data and algorithms shape AI models. | 1 | $1.6M | $1.4M–$1.6M | 2025 | ✓ |
| AI2050 | Schmidt Sciences allocated up to $5.5M to the 19 early-career fellows of the 2023 cohort (schmidtsciences.org) (up to $300k each over two years); an even share is ~$290k. Both figures are caps: no fellow got more than $300k, and the cohort total may be below $5.5M, so any one fellow may have got less than this share. | 5 +14 not counted | $1.4M | $290k–$300k | 2024 | ✓ |
| AI2050 | Schmidt Sciences announced up to $12M for the 2024 cohort (schmidtsciences.org) of 25 (5 senior, 20 early-career). Splitting it with senior fellows at 4x gives ~$300k per early-career fellow (5×1.2M + 20×300k = $12M). Since $12M is a cap ('up to'), the cohort may receive less in total, and any one fellow may get more or less than this share. | 4 +16 not counted | $1.2M | 2024 | ✓ | |
| Future of Life Institute | AI-safety share (80%, range 60–95%) of the 2024 grant to the Federation of American Scientists: FLI's grant page (futureoflife.org) says it is ‘To investigate the implications of artificial intelligence on global risk’, on ‘the nexus between AI and’ nuclear deterrence, bioengineering, autonomy and lethality, and cyber security (FAS announcement (fas.org)). The rest allows for general security-policy work in the same programme; no budget split is published. | 1 | $1.2M | $900k–$1.4M | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (85%, range 65–100%): Pivotal calls its biodefense fellowship ‘our first dedicated biosecurity programme, alongside our existing AI safety fellowship’ (pivotal-research.org), and 34 of the 35 outputs on its research page (pivotal-research.org) are AI safety or governance. No count of 2024 fellows by area is published. | 2 | $1.1M | $257k–$866k | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (90%, range 75–100%): Successif helps people ‘Build an impactful career in AI risk reduction’ (successif.org). | 1 | $1M | $862k–$1.1M | 2024 | ✓ |
| AI2050 | AI2050 allocated up to $7M to the seven 2023 senior fellows (srinstitute.utoronto.ca); an even share is ~$1M each. Since $7M is a cap ('up to'), the cohort may receive less in total, and any one fellow may get more or less than this share. | 1 +6 not counted | $1M | 2023 | ✓ | |
| Global Catastrophic Risks Opportunities | AI-safety share (15%, range 5–25%): GPI researched global priorities broadly, naming ‘the ethics of the long-term future, … the risks and opportunities of artificial intelligence, … existential risk, animal welfare, philanthropic decision-making’ among its topics on its homepage (globalprioritiesinstitute.org); 4 of the 23 GPI working papers dated 2023–24 on its publications list there are mainly about AI risk. No split of effort or budget by topic is published. | 2 | $838k | $125k–$770k | 2023 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (80%, range 60–95%): the grant funds the Tarbell Fellowship, and Tarbell says ‘We only support fellows interested in covering artificial intelligence’ (tarbellcenter.org). The 2023 grant text says ‘emerging technologies’ and also covers Training for Good’s operating costs, hence the range. | 1 | $799k | $599k–$949k | 2023 | ✓ |
| AI Safety Tactical Opportunities Fund | Not sourced, so left out of totals. AISTOF publishes no yearly figures: its Manifund page says only that it has raised more than $30M and made over 150 grants since September 2023. The 2024–2025 'Unknown Recipients' rows are Trace's assumed round totals (trace.manifund.org) ($5M, $15M, $25M) less itemised grants. The AI Safety Funder Bulletin's (manifund.org) ~$15M for 2025 is taken from Trace, and its $40M expected for 2026 rests on private communications. | 2 | $770k | $358k–$412k | 2024–2025 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (35%, range 20–50%): Consultants for Impact (consultantsforimpact.org) names three pressing problems it steers consultants to, ‘poverty, pandemics, risks from AI’, so AI is taken as about a third. No split of advising or placements by area is published. | 3 | $733k | $64k–$699k | 2023–2024 | ✓ |
| Musk Foundation | CSET's donor lists are cumulative: this funder moved from the $1M–$6M band (2019–23) to the $1M–$9,999,999 band (to 2024), so the 2024 row counts only the increase: the difference of the bands' geometric midpoints, likely $0–$9M. source | 1 | $713k | $713k–$9M | 2024 | ✓ |
| NobleReach Foundation | CSET's donor lists are cumulative: this funder moved from the $1M–$6M band (2019–23) to the $1M–$9,999,999 band (to 2024), so the 2024 row counts only the increase: the difference of the bands' geometric midpoints, likely $0–$9M. source | 1 | $713k | $713k–$9M | 2024 | ✓ |
| Diller–von Furstenberg Family Foundation | CSET's 2023 report ‘CSET at Five’ lists this funder in its $500,000–$999,999 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(500k × 1M) ≈ $707k. source | 1 | $707k | $500k–$1000k | ✓ | |
| Patrick J. McGovern Foundation | CSET's 2023 report ‘CSET at Five’ lists this funder in its $500,000–$999,999 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(500k × 1M) ≈ $707k. source | 1 | $707k | $500k–$1000k | ✓ | |
| Global Catastrophic Risks Opportunities | AI-safety share (65%, range 50–85%): the grants name workshops on AI and biotechnology risks, and GCP describes its workshops as ‘on the risks from advanced AI’ (globalchallengesproject.org), with biosecurity one of several topic workshops. No count of AI vs biosecurity workshops is published. | 2 | $609k | $221k–$422k | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (60%, range 40–80%): Meridian is ‘focused on AI safety, biosecurity, frontier-risk policy, and institutional design’ (meridiancambridge.org), and two of its three field hubs (Cambridge AI Safety Hub, UK AI Forum) are AI. No split of space or budget is published. | 1 | $574k | $383k–$766k | 2024 | ✓ |
| Survival and Flourishing Fund | AI-safety share (40%, range 25–55%) of 2025 general support to SecureBio: it lists two current projects, Detection and AI and Biotechnology Risks (securebio.org) (evaluations of whether AI systems could add to catastrophic biological risks), and about 23 of the roughly 62 people on its team page (securebio.org) are tagged AI against 28 Detection. No budget split is published. | 2 | $403k | $64k–$415k | 2024–2025 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (10%, range 5–25%): Non-Trivial (non-trivial.org) is for young people ‘solving the world’s most pressing problems’ across many areas, and 3 of the 31 example fellow projects on its homepage (non-trivial.org) are tagged Artificial Intelligence. No split of fellows or content by area is published. | 4 | $350k | $16k–$316k | 2022–2025 | ✓ |
| Alfred P. Sloan Foundation | CSET's 2023 report ‘CSET at Five’ lists this funder in its $200,000–$499,999 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(200k × 500k) ≈ $316k. source | 1 | $316k | $200k–$500k | ✓ | |
| NobleReach Foundation | CSET's 2025 annual report lists this funder among those who gave $100,000–$999,999 in 2025, with no exact amount. Set at the band's geometric midpoint, √(100k × 1M) ≈ $316k. source | 1 | $316k | 2025 | ✓ | |
| Rob Granieri Fund | CSET's 2023 report ‘CSET at Five’ lists this funder in its $200,000–$499,999 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(200k × 500k) ≈ $316k. source | 1 | $316k | $200k–$500k | ✓ | |
| Rockefeller Foundation | CSET's 2023 report ‘CSET at Five’ lists this funder in its $200,000–$499,999 donor band; the list seems to cover giving since 2019 (Open Philanthropy is at $100M+). Set at the band's geometric midpoint, √(200k × 500k) ≈ $316k. source | 1 | $316k | $200k–$500k | ✓ | |
| William and Flora Hewlett Foundation | CSET's 2025 annual report lists this funder among those who gave $100,000–$999,999 in 2025, with no exact amount. Set at the band's geometric midpoint, √(100k × 1M) ≈ $316k. source | 1 | $316k | 2025 | ✓ | |
| Coefficient Giving | AI-safety share (20%, range 10–35%) of general support to FABRIC: of its camps (ESPR, PAIR, ASPR, WARP; fabric.camp (fabric.camp)) only PAIR is about AI, and its curriculum (pair.camp) covers ‘some of the ethical and safety aspects’ of AI; the others teach applied rationality. No split by camp or budget is published. | 1 | $287k | $144k–$503k | 2024 | ✓ |
| Future of Life Institute | AI-safety share (25%, range 15–40%) of the 2022 grant to CSER: FLI's grant page (futureoflife.org) says CSER ‘studies existential risks’ and lists its year's work on climate, volcanic eruptions, pandemic preparedness, nuclear risk and AI/AGI, one of five areas. CSER publishes no split by area. | 1 | $250k | $150k–$400k | 2022 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (10%, range 0–20%) of a two-year 2022 grant: GPI researched global priorities broadly (its homepage (globalprioritiesinstitute.org) lists the long-term future, existential risk, animal welfare, philanthropy and AI risks among its topics). On its publications list none of the 4 GPI working papers dated 2023 is about AI and 4 of the 19 dated 2024 are mainly about AI risk, so the share is set below the 15% used for its 2023 grants. No split of effort or budget by topic is published. | 1 | $230k | $230k–$459k | 2022 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (35%, range 20–60%): the grant text (coefficientgiving.org) names three content areas (AI safety, global catastrophic risks, effective altruism) and gives no split, so AI safety is taken as about a third. | 1 | $229k | $131k–$392k | 2025 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (20%, range 10–35%): of FABRIC's camps (ESPR, PAIR, ASPR, WARP; fabric.camp (fabric.camp)) only PAIR is about AI, and its curriculum (pair.camp) covers ‘some of the ethical and safety aspects’ of AI; the others teach applied rationality. No split by camp or budget is published. | 2 | $216k | $50k–$204k | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (50%, range 30–70%): the grant text (coefficientgiving.org) says it supports ‘effective altruism and AI safety groups’ at London universities, so the AI safety groups are taken as about half. | 1 | $203k | $122k–$284k | 2023 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (80%, range 60–95%): the grant text (coefficientgiving.org) says the high-school courses teach neural networks ‘within the framework of AI safety’; the rest allows for general machine-learning teaching. | 1 | $200k | $150k–$238k | 2021 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (60%, range 35–85%) of the 2023 summer fellowship: 34 of the 35 fellow outputs on Pivotal's research page (pivotal-research.org) are AI safety or governance, and its only 2023-dated output is AI governance. The 2023 fellowship ran as the Swiss Existential Risk Initiative and its 2023 cause areas aren't published, so the share is set below the 85% used for 2024. | 1 | $172k | $100k–$244k | 2023 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (25%, range 15–40%): SERI works on five risk domains (seri.stanford.edu), one of them advanced AI (the others pandemics, nuclear, autonomous weapons, extreme climate change), and one of its six named courses (seri.stanford.edu) is AI Alignment. No split by risk area is published. | 1 | $158k | $95k–$252k | 2023 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (50%, range 35–65%): XLab's fellowship (xrisk.uchicago.edu) ‘focuses on AI strategy, security, and governance, and nuclear weapon security’, with AI and nuclear its two primary risk areas (xrisk.uchicago.edu); about 36 of the 57 fellows listed (xrisk.uchicago.edu) for 2022–25 have AI projects, some not safety-related. No count by area is published. | 1 | $151k | $106k–$196k | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (65%, range 50–85%): the grant text (coefficientgiving.org) says the mentorship programme covers ‘topics in AI safety, AI policy, and other areas’, so two of three named areas are AI. | 1 | $142k | $110k–$186k | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (60%, range 35–80%): the grants fund student programmes on existential risk 'such as AI alignment'. In June 2023 EffiSciences said (lesswrong.com) 'we are focusing on 3 topics: AI Safety, biorisks, and climate change' and reported results only for its AI safety unit (700 students reached). No split of effort or budget by topic is published. | 2 | $129k | $3k–$164k | 2023 | ✓ |
| Coefficient Giving | AI-safety share (50%, range 35–65%) of the 2025 XLab fellowship grant: XLab's fellowship (xrisk.uchicago.edu) ‘focuses on AI strategy, security, and governance, and nuclear weapon security’, with AI and nuclear its two primary risk areas (xrisk.uchicago.edu); about 36 of the 57 fellows listed for 2022–25 have AI projects, some not safety-related. No count by area is published. | 1 | $122k | $86k–$159k | 2025 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (10%, range 5–25%): Asterisk is a general-interest quarterly; 1 of its first 8 issues (asteriskmag.com) (2022–24) is AI-themed (issue 3, ‘AI’), and issues 4 and 7 carry one AI article each. No budget split by topic is published. | 1 | $113k | $57k–$284k | 2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (95%, range 85–100%): the grants pay for ML4Good bootcamps in Germany, and ML4Good (ml4good.org) says 'We train motivated people to contribute to AI Safety', naming no other cause area. This fund's later ML4Good grants (2023 via Dandelion Labs, 2025 direct) are already counted in full; the range allows for general machine-learning teaching. | 2 | $93k | $34k–$57k | 2023–2024 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (10%, range 3–20%): the 2022 conference's example day (futureforum.foundation) lists AI as one of five working-group tracks ('AI, Space, Biotech, Policy, others'), none of its three titled talks is about AI, and 2 of its 12 named speakers lead AI labs. No full programme is published, and AI talks need not have been about safety. | 1 | $65k | $20k–$131k | 2022 | ✓ |
| Science of Trustworthy AI | CSET's 2024 annual report lists Schmidt Sciences in its $15,000–$199,999 donor tier; Trace records the geometric mean of the tier, ~$55k. source | 1 | $55k | 2024 | ✓ | |
| Global Catastrophic Risks Opportunities | AI-safety share (5%, range 0–10%): of the 15 pieces in the magazine's archive (latecomermag.com), none has AI in its title and one (an interview on the future of intelligence) touches it; the editors' letters mention superintelligence. No budget split by topic is published. | 1 | $12k | $12k–$23k | 2022 | ✓ |
| Global Catastrophic Risks Opportunities | AI-safety share (20%, range 10–40%): Arb describes itself as a consultancy whose work 'spans forecasting, machine learning, theory, and policy' (arbresearch.com (arbresearch.com)); 3 of the 16 projects dated 2022–23 on its past-work list (arbresearch.com) are about AI safety and 6 touch AI at all. Arb says most of its work is private, so the public list is only a proxy; no split of effort by topic is published. | 1 | $5k | $3k–$10k | 2023 | ✓ |