Leif Wenar: AI needs people of good judgment, and effective altruists haven’t shown it

Leif Wenar, a philosophy professor at Stanford University, head of its ethics centre and author of “Blood Oil”, has written a sharp critique of the effective altruism (EA) movement in The Economist’s “By Invitation” column, arguing that what artificial intelligence needs is good judgment, not moral ambition. The piece ran on page 17 of the October 10, 2026 UK print edition under the headline “AI needs people of good judgment. Effective altruists haven’t shown it”, and online on October 7 as “Can effective altruists grow up?”.

Wenar’s main arguments

“What’s good isn’t new.” In Wenar’s view, what is good in EA isn’t new: compassion and equality were taught long ago by the Buddha, Jesus, Mo Tzu and the early British utilitarians. And what is new in the movement, he argues, isn’t good. Good judgment matters more than moral ambition.

Shaky sourcing. Wenar criticises William MacAskill’s “What We Owe The Future” for calling a climate charity “extremely cost-effective” on the strength of weak footnotes.

GiveWell’s numbers. GiveWell says it “will save over 340,000 lives.” Wenar argues that the figure leaves out harms well known to development economists, including actual deaths. He cites a top-rated cash charity in northern Nigeria where a bandit looking for the cash killed two people, and says those deaths are not counted as “minus two.”

Accountability. According to Wenar, GiveWell’s “lives saved” claim has not been peer-reviewed, and grant recipients have few rights to redress.

The SBF lesson. Wenar notes that Sam Bankman-Fried, who most fully lived out EA’s credo, is serving 25 years for fraud. Drawing on Tyler Cowen, he says the EA leaders around him failed to see the risks. Students now find EA “a bit cringeworthy,” he writes.

Not all bad. Wenar grants that funders now talk more like reasonable adults, and that money is flowing to basic science and public health.

The AI concern. Wenar’s real worry is people from EA building AI, especially at Anthropic. As he characterises it, Anthropic’s recent safety problems are being investigated by groups with close ties to the boss’s EA friends and family, and the main author of the moral “constitution” of Anthropic’s Claude has long involvement in EA. He sees few external checks on their choices.

His advice. Wenar urges would-be helpers to live among the people they want to help, gain deep expertise and acquire local life experience. EA’s final legacy, he concludes, may be measured in human lives, “just not in ways that they ever imagined.”

Context

• Effective altruism (EA): a movement, begun in Oxford in the late 2000s, that urges people to direct donations and careers to what evidence and calculation suggest will do the most good.
• GiveWell: a US non-profit that evaluates charities’ cost-effectiveness and recommends where to give. Its website says it “will save over 340,000 lives.”
• Sam Bankman-Fried (SBF): founder of the FTX crypto exchange. A US federal judge sentenced him to 25 years in prison for fraud and conspiracy on March 28, 2024 (US Department of Justice).
• Anthropic: the US AI company that develops the Claude models.

Editor’s note

Wenar’s critique is strongest on accountability and outside scrutiny. That estimates affecting millions of lives should be checked by independent researchers, and that recipients should have a voice, is a widely accepted principle in development work. His point that spreadsheets cannot replace real-world conditions and local knowledge is also valuable.

Some of his arguments are contested, however. GiveWell publishes its cost-effectiveness models and calculations in detail on its website, so it is not fair to call it opaque. EA defenders responded quickly on the EA Forum and elsewhere, arguing that people in EA warned about AI risk years before others did, and that a few overstated statistics and the SBF scandal do not prove that the whole movement lacks judgment. The passage on Anthropic is Wenar’s own characterisation, and we found no direct response from Anthropic to his column. The company did publish a report on October 9 describing unintended model actions seen in testing and internal use, and the measures it has taken.

For AI governance, the lesson is that independent oversight, transparency and accountability matter more than trust in any one group’s ethical outlook. For Mongolia, that means assessing foreign AI companies’ ethics principles and “constitutions” against Mongolian language, culture, law and social conditions before adopting them, rather than taking them on trust. When AI is introduced into government, independent audits, ways for citizens to complain and the involvement of local experts are essential. Wenar’s “good judgment”, in Mongolia’s case, means bringing people with local knowledge into decision-making.

Source: The Economist, “By Invitation”, Leif Wenar, October 10, 2026 (print edition, p.17) https://www.economist.com/by-invitation/2026/10/07/can-effective-altruists-grow-up