Some of Anthropic's earliest employees are reportedly considering buying land in remote parts of the United States as a place to relocate if advanced AI systems go wrong. The account comes from The Wall Street Journal and describes contingency planning rather than a company program or an imminent move.
The discussions fit a longer history inside the community that produced Anthropic and parts of OpenAI. Early company dinners reportedly included scenarios in which researchers might continue work in a shielded government facility. Several early hires also had ties to Effective Altruism, a movement whose members have spent years debating low-probability but catastrophic risks.
Those connections help explain why existential-risk arguments have had unusual influence in frontier AI labs. They do not establish that a disaster is likely, and the report does not identify how many employees are considering a purchase or whether anyone has completed one.
The practical news is therefore about the mindset around frontier-model development: some people closest to the work are treating extreme failure scenarios seriously enough to discuss personal preparations. The evidence remains secondhand and limited to an unspecified group of longtime employees.