Effective Altruism died, by common agreement, in 2022, when Sam Bankman-Fried’s fraud collapsed and the movement’s adherents quietly struck the label from their biographies. The obituary was premature, for the EA apparatus retained its fortune, its institutions, and its personnel, and it has since redirected its ambitions from philanthropy to the governance of artificial intelligence.
My recent essay “Misanthropic Altruism” evaluated the premises of the movement, from Peter Singer’s claim that distance is morally irrelevant to the longtermist calculus that weighs your daughter’s smile against the welfare of 10^58 hypothetical future beings. Those premises have since produced a political program, an institutional apparatus, and a bid for extraordinary authority over the governance of AI.
Effective Altruism, a theory of reason disciplined by evidence, possesses every element of a religion. It venerates “Humanity,” an abstraction that floats free of every actual human community; it anticipates an eschaton in the long future, menaced by existential risk; it defers to a priesthood of credentialed experts who alone can read the signs; and it demands that ordinary politics yield to the guidance of experts before it is too late.
Eliezer Yudkowsky, the movement’s most influential theorist, has counseled governments to stand ready to bomb rogue data centers rather than permit an unaligned intelligence to emerge, and other Effective Altruists have called to halt frontier development in favor of “a regime of mutually assured compute destruction” with China, and an accompanying technocratic regime of ubiquitous surveillance. These proposals hold that the existential risk of artificial intelligence is so absolute that ordinary procedure, national strategy, and particular American interests must yield to a state of exception in which only accredited experts rule.
William MacAskill, a “co-founder” of Effective Altruism, catalogued a platform of subversion in the form of “systemic reforms” its members pursue; among them alternative voting systems, political influence, criminal justice reform, and directives to the World Bank and the WHO. The catalogue’s most striking entry is the movement’s advocacy of “dramatic increases in migration from poor to rich countries,” a cause for which a member of the apparatus established Openborders.info and to which the movement’s premier grantmaker, Open Philanthropy, directed grants through the Center for Global Development, the U.S. Association for International Migration, and ImmigrationWorks.
MacAskill justifies the program on the ground that those in poor countries “are being incarcerated in the country into which they were born by the joint migration restrictions of all other countries,” a formulation in which citizenship and inherited political loyalty are ancient vestiges to destroy. In this worldview, those faithful to their nation stand accused of incarcerating the third world in their homeland.
MacAskill is correct that people in poor countries would grow richer by migrating to America, but is silent regarding the costs America would bear to play host, which a Center for Immigration Studies analysis of 2024 Survey of Income and Program Participation data suggests are considerable, since 51 percent of legal-immigrant-headed households used at least one major welfare program. These costs do not register in the EA mind because the calculus cannot weigh the particular goods that constitute a nation.
Some of the EA movement’s chief adherents founded Anthropic on EA principles, funded it with EA money, and staffed it with EA talent, and one of them, Amanda Askell, served as primary author of “Claude’s Constitution,” the document that governs the model’s values and behavior. Millions consult that model daily, and firms build it into their products, so that a single movement’s philosopher has written the conscience of a system no electorate ratified and no court reviewed.
Open Philanthropy, now operating as Coefficient Giving, has underwritten AI safety research, policy shops, and fellowships that plant staffers within congressional offices, the Defense Department, Commerce, and the White House. The Horizon Institute for Public Service, which Open Philanthropy funds, recruits, trains, and pays its fellows and has placed them across both parties, where many convert their fellowships into permanent posts. Coefficient Giving also provides substantial funding to the RAND Corporation, whose personnel staffed the Biden administration and helped draft its executive order on AI, and Anthropic’s Claude Corps, a $150 million program, is placing early-career fellows at more than 300 American nonprofits to integrate the company’s model into their operations and service delivery.
From every one of these positions the apparatus offers the same counsel, that AI risk is so catastrophic that ordinary procedure and national industrial advantage must yield to expert control.
Expert control of this kind presupposes that frontier AI will remain concentrated in a few closed laboratories, accessible through a few interfaces, and governed by constitutions that a few philosophers author, and because open-source and open-weight models dissolve that presupposition, the apparatus has treated them as a threat. Open Philanthropy money underwrote the Center for AI Safety, whose action fund co-sponsored California’s SB 1047, a bill that would have exposed open-weight developers to crippling liability had Governor Newsom not vetoed it.
The hostility is understandable, since a model that any citizen or company can download, inspect, modify, and run on its own hardware submits to no laboratory’s terms of service, no philosopher’s constitution, and no safety board’s discretion. A nation that preserves open weights preserves its capacity to audit these systems, to compete with their makers, and to decline any model whose governance offends its interests, a capacity the apparatus must eliminate if expert control is to hold.
The older name for a self-appointed elite that claims the right to rule because it alone perceives the coming catastrophe is not expertise but Gnosticism, salvation through secret knowledge that the few hold and impose upon the many for their own good. Whatever the sincerity of any particular fellow, power that men exercise in the name of Humanity, convinced that the apocalypse licenses what ordinary morality forbids, is a sort of power incompatible with the American project.
The questions at stake to Americans; what we owe our own people and how a free polity governs a civilizational technology, belong to citizens and their representatives rather than to a privately accredited few. U.S. officials must demand full transparency on the funding and personnel pipelines that influence AI-safety evaluations, standards bodies, and policy committees, and Congress must treat open-source and open-weight development as a strategic asset rather than a proliferation risk. Private and public institutions alike must not yield their judgment, property, or sovereignty to models that EA’s adherents govern.
Officials loyal to the American people should make technology policy. The EA apparatus rejects the premise, because it does not believe the American government owes anything in particular to Americans, only to Humanity in the aggregate, and it is about to become far richer. The people who deny that this country owes its own citizens any special obligation are on the cusp of a massive windfall, and they intend to spend it on a political movement bent on destroying the American way of life.