WSJ reports some early Anthropic staff are buying remote US land as an AI failure escape plan
The Wall Street Journal reported on September 26 that some early Anthropic employees are discussing buying land in remote US areas as a contingency if AI goes badly wrong.
IT Home on September 27 cited a Wall Street Journal report that several early Anthropic employees are quietly preparing a contingency plan that includes buying land in remote parts of the United States as a place to retreat if artificial intelligence turns on humanity. The plan has been shared with industry peers over the past few weeks and traces back to similar "Manhattan Project"-style scenarios discussed at company gatherings in Anthropic’s earliest days.
According to early Anthropic employees who have since left, the company’s early gatherings included discussion of a scenario in which the team might be called up by the government to a remote, EMP-shielded military base in the desert to continue working on frontier AI in secret. Staff also run end-of-the-world scenarios in private Slack channels, and a slogan — "things will not return to normal" — has reportedly been printed on stickers stuck to their laptops.
Asked about the culture, an Anthropic spokesperson said the company now has more than 3,500 employees and that internal technical discussions span a wide range of views. The report did not name anyone or give a headcount, but framed the plans as gradually shifting from jokes into long-term contingency work. Anthropic this week also separately unveiled Claude’s nine-loop N=4 super-Yang-Mills amplitude result.