Anthropic: Veterans Consider Remote Lands in Response to AI Risk

Some of the longest-serving employees at Anthropic are exploring the purchase of remote land in the United States to relocate if AI were to become uncontrollable. This initiative is part of an existential risk culture rooted in Effective Altruism, which has been associated with the early hires at Anthropic and OpenAI, and fueled by warnings from Eliezer Yudkowsky since the mid-2000s.

Roots Anchored in Existential Risk

The early recruits of Anthropic and OpenAI are linked to the subculture of Effective Altruism. Some come from a Bay Area network dubbed "doomers," which has spent years contemplating apocalyptic scenarios. Eliezer Yudkowsky raised alarms about the risk of uncontrollable AI in the mid-2000s and is regarded as an intellectual figure of this movement. Before AI became their primary concern, these groups were already interested in other catastrophic risks such as asteroid impacts or supervolcanoes.

Recently Discussed Plans for Isolated Lands

Some of the longest-serving employees at Anthropic are preparing contingency plans in case artificial intelligence becomes uncontrollable. In recent weeks, several of them have informed a colleague in the industry that they are considering purchasing land in remote areas of the United States. The goal would be to relocate there if AI "goes wrong," according to reports from the Wall Street Journal.

An Imaginary Precedent: The Shadow of the Manhattan Project

Former employees recall that during company dinners in the early days of Anthropic, a scenario inspired by the Manhattan Project was already being discussed. In this hypothesis, teams might be required to move to the desert to continue developing AI in a government facility shielded from electromagnetic effects.