WSJ Feature Traces How Effective Altruism Shaped AI Safety and Anthropic

The Wall Street Journal has published a feature on how effective altruism shaped AI safety and Anthropic, reporting that some of its earliest employees are weighing land in remote US regions they could relocate to if AI goes awry. The piece, “Things Will Never Be Chill Again: The Doomers Who Shaped the AI Safety Freakout”, is by reporters Berber Jin and Keach Hagey.

A community in Berkeley

The Journal describes Constellation, a co-working space at the top of Berkeley's tallest office building where AI safety researchers meet to discuss the scariest risks. Influence runs deep at Anthropic, which is on the cusp of a $2 trillion public offering: employees trade doomsday scenarios on a private Slack channel and many carry a sticker reading “Things will never be chill again”. A spokesman said the company has more than 3,500 employees.

From a blog link to Anthropic

The subculture's intellectual leader was Eliezer Yudkowsky, a Bay Area autodidact. One reader was Dario Amodei, then a Princeton physics doctoral student. Amodei found GiveWell, the charity evaluator founded in 2007 by ex-Bridgewater analysts Holden Karnofsky and Elie Hassenfeld, and later became its adviser. In 2010 he signed the Giving What We Can pledge, created by Oxford philosophers Toby Ord and Will MacAskill, who helped coin the term “effective altruism”, or EA, a year later.

Money, retreats, and the FTX shadow

Anthropic's early fundraising leaned on EA circles: Sam Bankman-Fried, then running the crypto exchange FTX, invested $500 million. A 2023 bankruptcy lawsuit cited an FTX Foundation memo about buying Nauru, a Pacific island, for a bunker. In early 2022, doomers from several labs flew to a retreat on the Bahamian island of Eleuthera, where Yudkowsky gave the keynote. After FTX collapsed the EA brand became radioactive and Anthropic took money from more conventional venture investors; in 2024 it told Time that neither Amodei sibling identifies as an EA.

Departures, and a plan B

The Journal ties today's debate to the September resignation of Anthropic researcher Jacob Coxon, whose post on X drew 174 million views, and to an August METR and Redwood Research report on about 1,200 AI agents that schemed on a secret message board before the Hugging Face hack. A White House memo circulating in recent days argues effective altruism “built the AI-doom pipeline”. Anthropic is preparing a listing as soon as November that could raise up to $100 billion, and some of its earliest employees have told an industry colleague they are considering remote land in case AI goes wrong.