Some Anthropic veterans reportedly buying remote land as AI fallback

By Splitfeed EditorialLast updated Sep 27, 2026, 13:10 UTC

A Wall Street Journal report published on September 27, 2026 describes a subset of Anthropic's longest-serving employees who are reportedly considering purchasing land in remote parts of the United States as a contingency if AI development produces catastrophic outcomes. The report links these employees to the Effective Altruism movement and to a Bay Area network described as having spent more than a decade modeling doomsday scenarios involving AI. No specific employees are named in the available sourcing, and no completed land purchases are confirmed.

The story matters because Anthropic is not an ordinary tech company. It was founded in 2021 by former OpenAI researchers specifically around the premise that advanced AI carries serious risks, and it has positioned AI safety as central to its mission and commercial identity. If even a portion of the company's most tenured staff are privately preparing for civilizational disruption, that raises pointed questions about the gap between the reassuring public language that AI companies use and the private assessments of the people building these systems. What remains genuinely in dispute is how representative these individuals are, whether land-buying consideration reflects a fringe cultural habit inherited from the EA community or something more widespread, and whether such behavior should be read as responsible contingency planning or as a signal that insiders believe the risks are more severe than publicly acknowledged.

Because no camp has published reactions to this story yet, their likely positions can only be anticipated. The Pro-AI camp would probably argue that a handful of employees with EA backgrounds engaging in personal contingency planning says nothing meaningful about AI risk at scale, and that conflating individual behavior with institutional belief is misleading. The Anti-AI camp would likely treat the story as revealing confirmation that those closest to the technology do not trust it, treating the land-buying consideration as an implicit admission that catastrophic outcomes are plausible. The Middle Ground camp would probably call for distinguishing between personal risk tolerance and evidence-based threat assessment, and caution against reading private lifestyle choices as technical or policy conclusions.

The more significant indicators to watch will be whether Anthropic responds officially, whether the Wall Street Journal publishes fuller sourcing with named employees or confirmed transactions, and whether this story prompts broader scrutiny of how AI safety researchers privately model the timelines and severity of the risks they study professionally.