FTX Future Fund

Theory of Change

FTX Future Fund was a longtermist philanthropic fund that operated for approximately 9 months (February-November 2022) before collapsing with the FTX bankruptcy. Its theory of change was direct from founder Nick Beckstead's PhD thesis, "On the Overwhelming Importance of Shaping the Far Future": because the far future contains vastly more potential value than the present, actions that reduce existential risk or improve humanity's long-run trajectory dominate expected value calculations. FF aimed to deploy capital at unprecedented speed and scale into AI safety, biosecurity, epistemic institutions, and EA field-building.

Beckstead, in a 2017 80K Hours podcast, explained the philosophy: "Almost all of the potential value is in the distant future... There are some things that we can do now, particularly in terms of understanding and mitigating potential global catastrophic risk, that have the potential to shape basically how large and good that future is."

FF's distinctive contribution was the regranting model: 100+ trusted individuals given authority to make grants on FF's behalf, an experiment in parallelizing philanthropic decision-making. This was paired with traditional staff-led grantmaking (large institutional grants) and an open call (applications from anyone).

What They Do

FF committed ~$160M in grants and investments across ~262 recipients in approximately 5 months of active grantmaking (March-August 2022). This occurred through three tracks:

Staff-led grants: ~$73M across 25 grants, averaging $2.9M. Largest: Longview Philanthropy ($15M), Centre for Effective Altruism ($13.94M), HelixNano ($10M pandemic vaccine), Atlas Fellowship ($5M talent pipeline), Ought/Elicit ($5M AI tools), Lightcone Infrastructure ($2M).
Regranting program: ~$31M across 168 grants via 100+ anonymous regrantors. Grants ranged from $30K individual research stipends to $1M+ for organizations like Manifold Markets. Funded AI safety interpretability research, career transition grants, forecasting platforms, Spanish-language EA translations, and NeurIPS existential risk prizes.
Open call: ~$26M across 69 grants. Notable: $6.5M to Center for AI Safety, $1.2M to SecureBio (Kevin Esvelt), $600K to UC Berkeley (Sergey Levine, AI deception research), $290K to AI Safety Camp, $200K to Anysphere (the company that became Cursor, now valued at ~$10B).

AI safety-specific grantmaking totaled ~$32M. A $6.6M grant to Redwood Research was committed but never paid.

Collapse (November 10-11, 2022): The entire FF team resigned en masse, citing "fundamental questions about the legitimacy and integrity of the business operations." FTX filed for bankruptcy the following day. An estimated $25M+ in pipeline grants were never honored. PhD students, postdocs, and individual researchers structured around FF funding were left stranded.

Clawback aftermath: The FTX bankruptcy estate pursued recovery from grantees. Effective Ventures returned $26.8M (100% of received funds, January 2024). CAIS faced a $6.5M clawback probe. CFAR was sued for $5M. The estate sought $71M total from philanthropic arms.

Key People

Nick Beckstead (CEO): PhD Philosophy from Rutgers, former Program Officer at Open Philanthropy, former Research Fellow at FHI. Left Open Phil for FTX Foundation in November 2021. Was personally warned about SBF's dishonesty and unethical behavior in 2018 by former Alameda employees; went on to lead SBF's philanthropy four years later. After collapse, left EV boards (August 2023). Now CEO/co-founder of Secure AI Project.

Notable dispersal pattern: Leopold Aschenbrenner (FF staff) went to OpenAI Superalignment (fired 2024), then founded $1.5B hedge fund. Avital Balwit (FF staff, also managed SBF-funded congressional campaign) became Chief of Staff to Dario Amodei at Anthropic. Ketan Ramakrishnan became a Yale Law professor. MacAskill (FF adviser) shifted to "post-AGI governance" research.

Team size: 5 core team members (Beckstead, Aschenbrenner, Balwit, MacAskill, Ramakrishnan). FTX Foundation board: Caroline Ellison, Gary Wang, Nishad Singh -- all three later convicted of fraud.

Money and Incentives

Sole funding source: Sam Bankman-Fried, via FTX/Alameda Research. All of FF's funding came from SBF and his associates (Ellison, Wang, Singh contributed to the FTX Foundation). As established at trial, the underlying money was substantially derived from misappropriated FTX customer deposits. The FF team claims -- and no evidence contradicts -- that they did not know the money was stolen.

The EA financing chain: Alameda Research was itself seeded by EA donors: Jaan Tallinn loaned $110M in Ether, Luke Ding invested $6M. Nishad Singh said: "This thing couldn't have taken off without EA. All the employees, all the funding -- everything was EA to start with." EA didn't just receive SBF's money -- it helped create SBF's money.

Concentration in EA-connected orgs: $36.5M went to organizations connected to MacAskill (FF adviser): $13.94M to CEA, $15M+ to Longview Philanthropy (where MacAskill sat on the advisory board). No public conflict-of-interest disclosures or recusal processes.

Parallel SBF spending: SBF separately invested $500M in Anthropic's Series B (April 2022, via Alameda, using customer funds -- 86% of the round). SBF also spent $40M+ on political campaigns and funded the Carrick Flynn congressional race (~$11M via EA-connected donors). Total SBF philanthropic/political spending dwarfed FF's $160M.

No financial disclosure: No 990 filings have been found for FTX Foundation. Officer compensation, administrative costs, and complete grant schedules remain undisclosed. The regranting program's 100+ regrantors were never publicly identified.

Ecosystem impact: FF's collapse was the single most disruptive event in AI safety funding history. Open Philanthropy provided emergency bridge funding ($753K for career transitions). 40% of EA funding had been directed to longtermist causes by 2022, and FF's collapse revealed the extreme concentration risk in the ecosystem's donor base.

What Others Say

The 2018 warnings (TIME, March 2023): Tara Mac Aulay and other former Alameda employees warned MacAskill, Beckstead, and Karnofsky in 2018 that SBF was "untrustworthy," engaged in "inappropriate sexual relationships with subordinates," refused standard business practices, and had been caught lying. A planning document for an Alameda "intervention" accused SBF of "gross negligence" and "willful and knowing violations." MacAskill dismissed the concerns as "rumor" and, according to one source, "basically took Sam's side" and "threatened Tara."

The structural critique (American Prospect): EA's "earn to give" framework makes the earning side unexamined. Longtermist expected value calculations -- multiplying tiny probabilities by astronomically large future populations -- naturally justify enormous risky bets. SBF's gambling philosophy was not an aberration from EA principles but a predictable product of them. "The consistent application of formal rationality to ethical problems -- if not disciplined by common sense, basic decency, and, above all, a structural analysis of power -- is apt to birth monsters."

Tyler Cowen: "Hardly anyone associated with Future Fund saw the existential risk to... Future Fund, even though they were as close to it as one could possibly be. I am thus skeptical about their ability to predict existential risk more generally."

Internal EA critics (New Statesman): Former CEA CEO Larissa Hesketh-Rowe: "When people believe their cause is especially important and that they are especially smart, it can be all too easy to stray into finding ways to justify behaviour that stems from normal human flaws." A former senior EA: "If you didn't investigate [the 2018 warnings] thoroughly you're being extremely foolish. And if you did, I think you would have known that there were reasons to at least be very concerned." On the movement's leaders going quiet: "These people were all happy to take talent, take status, while the times were good. And as soon as the times were bad, they decided to get on the first boat."

MacAskill's self-reflection (April 2024): Admitted EA was "too trusting" and "too reckless at times." Said the Alameda dispute was "a foreshadowing." Now says: "Looking forward, I'm going to be less likely to infer that, just because someone has sincere-seeming signals of being highly morally motivated... they will have moral integrity in other ways, too."

What's Absent

No 990 filings -- officer compensation, administrative costs, and complete financials are unknown for a funder that disbursed $130M+.
Anonymous regrantors -- 100+ individuals distributed $31M with no public accountability. Identities never disclosed.
No decision-making documentation -- evaluation criteria, rejection rates, and speed of decisions are undisclosed for a funder claiming rigorous consequentialist analysis.
No post-mortem from FF leadership -- Beckstead has not spoken publicly about what he would have done differently in grantmaking approach.
Incomplete clawback data -- the full list of unfulfilled grants, and which grantees faced hardship, is scattered and incomplete.
No conflict-of-interest policy -- for a funder giving $36.5M to organizations its own advisers founded, the absence of formal COI procedures is remarkable.