The private contingency
Effective Altruism, With an Escape Route
Some early Anthropic employees are reportedly considering remote land in case AI goes awry. It is an awkward footnote to the promise of a better future for everyone.

The people building AI like to talk about the future of humanity. Some of the people who helped build Anthropic are apparently also thinking about where to go if that future gets out of hand. It’s a useful distinction: humanity is a very large group, but a remote property has a much shorter guest list.
The Decoder, citing The Wall Street Journal, reports that some of Anthropic’s earliest employees are considering buying land in remote parts of the US in case AI “goes awry.” Considering, not buying: there is no settlement with a gate that says WE DID WARN YOU. Yet.
Former employees also recall a topic of conversation at company dinners in Anthropic’s early days: a Manhattan Project-style arrangement in which AI work might move to an electromagnetically shielded government facility in the desert. Most work dinners struggle to settle where to eat next time. This one apparently made room for where to build the future of civilization.
None of this is especially strange on its own. People who work on a risky technology worry about the risks. They make contingency plans. The comedy starts when the ambition gets big enough to encompass everyone on Earth, while the contingency plan gets small enough to fit on a deed.
In the public pitch, everyone has a stake in the future. In the private one, it might be worth checking the road access first.
Effective altruism was supposed to ask how to do the most good for the most people. Here’s a wonderfully awkward addition to the spreadsheet: what does land far from other people cost? In the public pitch, everyone has a stake in the future. In the private one, it might be worth checking the road access first.
A conversation over dinner isn’t a government project, and a property search isn’t proof of impending disaster. It is, however, a funny look at the limits of confidence. When people close to the work claim unusual insight into how humanity should build its future, you can be forgiven for noticing that some of them are also pricing an exit.
Imagerly.Ai · No. 05 · September 2026