Effective Altruism

Abstract

Effective altruism (EA) began as a question asked by Oxford philosophers and New York hedge-fund analysts around 2007: which charity saves the most lives per dollar? Within fifteen years it had become the moral philosophy of a generation of tech founders, moved billions of dollars into malaria nets and AI-safety research, helped fund OpenAI and Anthropic, and given Silicon Valley the word “longtermism.” Then its best-known donor, Sam Bankman-Fried, turned out to be running a fraud, and in November 2023 a board fight at OpenAI attached the label to a failed attempt to remove Sam Altman. This article covers the movement’s origins, its branches, its effect on the AI industry and its fall from fashion; for its place among Silicon Valley’s other schools of thought, see Silicon Valley Philosophies.

The Drowning Child

The movement’s founding argument is fifty years older than the movement. In “Famine, Affluence, and Morality” (1972) the Australian philosopher Peter Singer asked his readers to imagine walking past a shallow pond in which a child is drowning. Anyone would wade in, even at the cost of ruining their clothes. Singer argued that distance makes no moral difference: if you can prevent a death in Bengal at the cost of a new suit, you are obliged to. He returned to the argument in The Life You Can Save (2009).

What was missing was a way to tell which of the world’s charities actually prevented the deaths. In 2007 two analysts at the hedge fund Bridgewater Associates, Holden Karnofsky and Elie Hassenfeld, founded GiveWell to answer that question with the same tools they used to evaluate investments: randomized trials, cost-effectiveness models and published spreadsheets. Most charities they asked could not say what their programs achieved. A few could. GiveWell’s top recommendations settled on cheap, well-studied health interventions such as insecticide-treated bed nets distributed by the Against Malaria Foundation, for which its models have typically estimated an average cost of $3,000 to $5,500 per life saved.

Oxford and the Name

In Oxford the philosopher Toby Ord decided to give away most of his lifetime income. In 2009 he, Bernadette Young (then a physician in training) and the philosopher William MacAskill founded Giving What We Can, whose members pledge at least 10 percent of their income to the most effective charities. MacAskill and Benjamin Todd founded 80,000 Hours in 2011, named after the approximate number of hours in a working life, to advise graduates on high-impact careers.

The movement needed an umbrella organization and therefore a name. On 5 December 2011 a vote among about seventeen people in the Oxford group chose “Centre for Effective Altruism”; the shortlist of fifteen had included “Rational Altruist Community” and “Optimal Altruism Network.” The name was meant for the legal entity, but the phrase spread. Singer’s TED talk of 2013 and MacAskill’s Doing Good Better (2015) brought it to a general audience.

Earning to Give

80,000 Hours’ most famous and most disputed recommendation was earning to give: a talented graduate might do more good by taking a high-paying job in finance or technology and donating much of the salary than by working for a charity directly. The advice suited the people it reached. Effective altruism grew fastest among students at elite universities and young employees in quantitative finance and Silicon Valley, who were used to optimizing metrics and had money to give.

The money that defined the movement came from tech. Dustin Moskovitz, a co-founder of Facebook, and his wife Cari Tuna pledged most of their fortune through their foundation Good Ventures and partnered with GiveWell; the partnership became Open Philanthropy, which by the 2020s was one of the largest grant-makers in the world. It renamed itself Coefficient Giving in November 2025. GiveWell alone directed $397 million to its recommended programs in the twelve months to January 2025, about two-thirds of it to malaria programs.

Longtermism and AI

The second wave of EA moved its attention from present suffering to the far future. Nick Bostrom’s “Astronomical Waste” (2003) had argued that because the future could contain vastly more people than the present, even a small reduction in the risk of human extinction is worth more than almost anything else. MacAskill and Ord named the view longtermism around 2017. Ord’s The Precipice (2020) estimated the chance of an existential catastrophe this century at about one in six, with unaligned AI the largest single risk. MacAskill’s What We Owe the Future (2022) brought the argument to the bestseller lists, and Elon Musk tweeted that it was “a close match for my philosophy.”

Here EA merged with the rationalist community of Eliezer Yudkowsky, which had argued since the early 2000s that superintelligent AI was the main threat. The merger had direct effects on the AI industry:

In 2017 Open Philanthropy granted OpenAI $30 million, and Holden Karnofsky joined OpenAI’s board to help oversee its safety and governance work.
In 2021 a group of OpenAI researchers left to found Anthropic, with early money from EA-aligned investors including Jaan Tallinn and Moskovitz.
Open Philanthropy, the movement’s largest funder, gave about $336 million to AI safety between 2017 and 2023, by a community estimate probably more than any other funder in the world, and 80,000 Hours ranked AI safety as the top career priority.

Critics inside and outside the movement noted that longtermism shifted EA’s money from verifiable benefits (bed nets) toward speculative ones (research on systems that did not yet exist) whose results could not be measured, the opposite of the method GiveWell was founded on.

Sam Bankman-Fried

The movement’s most visible member was Sam Bankman-Fried. As an MIT student he met MacAskill, who suggested he earn to give rather than work directly on animal welfare. He went to the trading firm Jane Street, then founded the crypto trading firm Alameda Research and the exchange FTX, and talked about EA in interviews as the reason he wanted to be rich. In February 2022 his FTX Foundation launched the FTX Future Fund, led by EA figures including MacAskill, with plans to give away at least $100 million that year.

FTX collapsed in November 2022. Customer deposits had been used to cover Alameda’s losses, and Bankman-Fried was convicted of fraud in November 2023 and sentenced to 25 years in March 2024. The Future Fund’s team resigned on 10–11 November 2022, writing that “it looks like there are many committed grants that the Future Fund will be unable to honor.” MacAskill wrote: “if that goodwill laundered fraud, I am ashamed.”

The damage spread. Effective Ventures, the charity that housed the Centre for Effective Altruism, agreed in January 2024 to pay the FTX estate $26.8 million, the amount it had received from FTX and the FTX Foundation. It had also bought Wytham Abbey, a manor house near Oxford, in April 2022 for £14.9 million with grants from Open Philanthropy (about £17 million in all, covering purchase, renovation and running costs), to use as a conference venue. After FTX, critics cited the purchase as evidence that a movement preaching cost-effectiveness had bought itself a castle. The Abbey was put on sale in 2024 and sold in 2025 for £5.95 million.

The OpenAI Board

A year after FTX, effective altruism was blamed for a second crisis. On 17 November 2023 OpenAI’s board fired Sam Altman, saying he had not been “consistently candid.” Two of the four directors who voted to remove him, Helen Toner and Tasha McCauley, had ties to EA organizations; Toner had worked at Open Philanthropy. Within days nearly all of OpenAI’s staff threatened to leave, Altman was reinstated, and Toner and McCauley left the board. Whether EA motives drove the decision remains disputed, and later reporting pointed to concerns about Altman’s management reported by OpenAI executives. In public perception the episode fixed EA’s image as a faction that had tried and failed to take control of the most valuable AI company.

Branches

By the mid-2020s the movement had split along lines that had been visible from the start:

Branch	Priority	Representative institutions
Global health and development	Measurable lives saved now	GiveWell, Against Malaria Foundation, Giving What We Can
Animal welfare	Suffering of farmed animals	Animal Charity Evaluators, cage-free campaigns
Longtermism and existential risk	Survival of humanity’s future	80,000 Hours, Bostrom’s former FHI, Coefficient Giving’s AI programs
AI safety (builder wing)	Making frontier AI safe from inside the labs	Anthropic, safety teams at OpenAI and Google DeepMind
AI safety (pause wing)	Stopping frontier AI	MIRI, PauseAI

The branches disagree with each other about almost everything except the method of asking how much good a dollar or an hour can do. Giving What We Can reached 10,000 people who had taken its 10 percent pledge in August 2025, and the global-health branch kept growing quietly while the AI branches argued in public.

Dead End: The Movement as a Brand

The effective altruists’ method, evaluating charities by evidence and cost per outcome, has outlived the movement’s reputation. GiveWell’s top charities still receive hundreds of millions a year. What did not survive was EA as a badge that tech founders wore in public. Before November 2022 Bankman-Fried presented himself as its model donor and Musk endorsed its longtermist bestseller; afterwards founders who shared its ideas about AI risk mostly stopped using the label. A common critique after 2022 did not concern bed nets. It was that a philosophy which tells individuals to maximize expected value can justify almost anything once the numbers are large enough, including, in Bankman-Fried’s case, gambling with other people’s money.