Anthropic veterans eye remote US land if AI goes awry: WSJ

The report places the land-buying talk inside a longer-running habit of worst-case planning at Anthropic. Former employees recall that even at company dinners during the company's early days, people discussed a Manhattan Project-style scenario in which they might be asked to move to the desert and keep building AI inside an electromagnetically shielded government facility. That context matters because the same article links the earliest hires at Anthropic and OpenAI to the Effective Altruism subculture and to a Bay Area network of "doomers" who have spent years gaming out end-of-the-world scenarios.

The intellectual lineage the WSJ traces runs through Eliezer Yudkowsky, who sounded the alarm about uncontrolled AI in the mid-2000s and became the godfather of that subculture. Before AI topped the threat list, the same group obsessed over catastrophic risks ranging from asteroid impacts to supervolcanoes. Seen against that history, the reported land purchases look less like a break and more like the latest expression of a worldview already embedded in the labs.

What the outcome hinges on is whether talk becomes action. The report describes consideration, not completed purchases, so the significance may rest on whether these plans stay private contingency thinking or become visible commitments that shape how the labs are perceived by recruits, investors and regulators. For the people deciding where to work in frontier AI, the report offers one more data point about how the earliest insiders weigh the risks they are building around.