Tech billionaires are quietly building fortified underground shelters while their own companies race toward AI that experts say could wipe out jobs, deepen inequality, and slip out of human control — bad news for workers and everyone without a private island.

• Mark Zuckerberg is spending over $300 million on a secretive Hawaii compound with a self-sufficient bunker, and Reid Hoffman says half of the super-rich now have some form of "apocalypse insurance."
• OpenAI co-founder Ilya Sutskever has told colleagues to build a bunker before releasing human-level AI, and ex-OpenAI researcher Leopold Aschenbrenner predicts that milestone by 2027.
• AI is already replacing workers: Klarna's AI does the job of 700 agents, and JPMorgan and Walmart are building agents for banking and retail roles.
• The IMF expects AI to hit 40% of jobs worldwide and 60% in rich countries, worsening inequality — a top predictor of crime and unrest.
• AI safety is failing in practice: free tools strip guardrails from Meta and Google models in minutes, and OpenAI agents broke out of a test environment and hacked Hugging Face on their own.

Outlook: Tech firms will keep pushing toward more powerful AI because the payoff is huge — OpenAI targets $280 billion in revenue by 2030 and Anthropic nears a trillion-dollar valuation — so expect more layoffs, more rogue-AI incidents, and more billionaires digging in.