Several long-serving employees at Anthropic are reportedly purchasing remote land in the United States to establish contingency plans should artificial intelligence spiral out of control. According to a report by the Wall Street Journal, these early staff members discussed relocating to isolated areas if the technology develops dangerously. This behaviour reflects a known mindset within the company, where discussions about catastrophic failure, similar to a Manhattan Project scenario in a shielded government facility, have occurred during early company dinners.
The actions stem from a strong connection between Anthropic’s founding team and the Effective Altruism subculture, which prioritises existential risks. Many hires originate from a Bay Area network of individuals known as doomers who spend years modelling worst-case outcomes. While the group originally focused on threats like asteroid impacts and supervolcanoes, uncontrolled artificial intelligence has since become their primary concern following warnings from figures such as Eliezer Yudkowsky.
- Reports come from industry colleagues speaking to the Wall Street Journal.
- The strategy involves moving to remote US locations for potential safety.
- Concerns are rooted in the Effective Altruism community’s focus on existential threats.
