AI Is Making the Prepper Movement Mainstream

The conference-call pods at East Room, the co-working space I frequent, are almost always busy — and hardly soundproof. Walking by, you can pick up fragments of go-to-market plans, pipeline reviews, or all-hands updates. Or, if you happened to overhear my conversation with Jérémie Harris, AI apocalypse scenarios. 

I got on the phone with Harris just days after Jacob Coxon, a former safety researcher at Anthropic, tweeted that AI has a 10% chance of wiping out humanity in the next 10 years. Harris is the co-founder and CEO of Gladstone AI, an AI safety company focused on catastrophic risks from advanced AI, and has spent the past two years doing security work for some of the largest AI cluster builds in the world — specifically, figuring out how to prevent countries like China and Russia from sabotage or espionage against those facilities. He’s also Canadian and met with more than 30 members of Parliament about AI risk in 2020 and 2021.

Through his work, Harris has rubbed elbows not only with Silicon Valley insiders but also members of the intelligence and special operations communities, including SEAL Team Six. “Many of them have well-developed ‘bug-out’ spreadsheets — lists of gear, staging plans, timing, and so on,” he says. (In prepper speak, bug-out refers to grabbing a go-bag and fleeing; bug-in refers to staying in place and riding it out.) “People are waking up to this and thinking, ‘I should probably put myself and my family in a position to weather a storm.’” What kind of storm? “A Hugging Face incident on steroids,” he says. In other words: an agent swarm that, whether by going rogue or being deliberately weaponized, takes out part of the power grid, water systems, or GPS.