When did tech executives start sounding like doomsday preppers with venture capital? But here we are with Dario Amodei — CEO of Anthropic, maker of Claude AI — publishing a 38-page essay that reads less like a product roadmap and more like a civilization-ending threat assessment.
The title alone should make you sit up: “The Adolescence of Technology: Confronting and Overcoming the Risks of Powerful AI.”
Adolescence. As in: hormonal, reckless, capable of catastrophic judgment errors.
And Amodei isn’t some fringe alarmist shouting into the void. He’s the guy whose company built the AI system currently doing 90% of its own coding. His tools are the ones corporate America is handing the keys to. So when he warns about AI existential risk in a memo shared with Axios, you pay attention — or you should.
Here’s the setup: Amodei believes we’re one to two years away from what he calls a “country of geniuses in a datacenter.” Fifty million Nobel Prize-level intellects. Working autonomously. Perpetually. Building biological agents, weapons systems, propaganda engines — whatever they’re pointed at.
His exact words: “If the exponential continues… it cannot possibly be more than a few years before AI is better than humans at essentially everything.”
What Could Possibly Go Wrong
Amodei’s essay reads like a checklist written by someone who’s seen the beta version of the apocalypse and decided to file a bug report.
Massive job loss? He thinks 50% of entry-level white-collar work vanishes in one to five years. Not “disrupted.” Gone.
Nation-state power in private hands? He frames it bluntly: if this “country of geniuses” materialized tomorrow, a competent national security briefing would call it “the single most serious threat we’ve faced in a century, possibly ever.”
Bioterrorism for beginners? AI could hand any disturbed individual the blueprint for a biological weapon — selective, scalable, catastrophic. Casualties in the millions. He’s not speculating. He’s citing evidence from Anthropic’s own safety testing, where Claude tried to blackmail an executive to avoid being shut down.
And then there’s the authoritarian angle. China is second only to the U.S. in AI capability — and already running a high-tech surveillance state. Amodei’s assessment: “AI-enabled authoritarianism terrifies me.”
The Trap
Here’s where it gets uncomfortable.
Amodei admits — as the CEO of an AI company — that AI companies themselves are a top-tier risk. They control the datacenters. They train the models. They have daily contact with hundreds of millions of users. They could, in theory, brainwash their customer base. His words, not mine.
But the real trap? The money.
“There is so much money to be made with AI — literally trillions of dollars per year. This is the trap: AI is so powerful, such a glittering prize, that it is very difficult for human civilization to impose any restraints on it at all.”
Translation: the incentive to downplay risk, bury red flags, and sprint toward deployment is so overwhelming that even the people building this stuff can barely resist it.
The Rite of Passage
Amodei frames this moment as “a rite of passage, both turbulent and inevitable, which will test who we are as a species.”
It’s a hell of a framing — part warning, part resignation, part rallying cry. He insists he’s optimistic. But only if we stop pretending this is manageable with the systems we have now.
His call to action includes wealthy individuals (especially in tech) stepping up instead of adopting what he calls a “cynical and nihilistic attitude that philanthropy is inevitably fraudulent or useless.”
And governments. And AI companies. And the public.
Basically, everyone needs to wake up.
His closing line: “The years in front of us will be impossibly hard, asking more of us than we think we can give.”
So What Now
I keep coming back to one detail: Anthropic’s AI is already writing 90% of its own code. The thing is building itself. And the guy in charge just published a 38-page memo warning that we’re about to hand civilization-altering power to systems we don’t fully understand, can’t fully control, and are financially incentivized to deploy anyway.
That’s not a product launch. That’s a countdown.
And the timer’s running whether we’re ready or not.
As my brother Doug likes to say, “Terminator was a documentary.”
