Anthropic Staff Weigh Remote Land Purchases as AI-Risk Fallout Plans, WSJ Reports

Some senior employees at AI safety company Anthropic are preparing contingency plans for a scenario in which artificial intelligence systems escape human control, including buying land and building hideouts in remote areas of the United States, The Wall Street Journal reported on Sept. 26, 2026. Early employees have told industry peers in recent weeks they intend to acquire property in isolated regions so their families could relocate if AI turns against humanity, the report said. The discussions echo a long-running internal culture: former staff recalled early company dinners that imagined a "Manhattan Project-style" scenario in which they would be conscripted to continue AI research inside an electromagnetic-pulse-shielded desert base. Employees have also debated doomsday scenarios on private Slack channels and printed a motto as a sticker on laptop cases: "Nothing will ever be normal again." Anthropic spokesperson said the company, which has more than 3,500 employees, holds highly diverse internal views on technology.