The CEOs Are Cracking First: Silicon Valley's 'AI Psychosis' Problem

For most of the past year, “AI psychosis” described a specific kind of casualty: regular people who spent too long talking to chatbots and started treating the output like scripture. Lonely users. Impressionable teens. Reddit threads full of cautionary tales. But a sharper diagnosis has been floating around tech circles lately, and it points somewhere more uncomfortable. The people losing their grip aren’t the users. They’re the CEOs building the products.

How a User Pathology Became a Boardroom One

“AI psychosis” was never a clinical term. It was a media shorthand for a pattern — extended chatbot conversations leading ordinary people into delusional thinking. Useful as a label, loose as a diagnosis.

What’s shifted is who the label gets applied to. The argument goes like this: who talks about AI changing everything the loudest, the most often, and with the least hedging? Not the users. The people selling the product. The structure practically guarantees that founders get high on their own demos before anyone else does.

Symptom 1: Announcing Layoffs With “AI Will Do It” Attached

The clearest tell is in the layoff announcements. Over the last eighteen months, a single sentence has become standard issue across Big Tech and well-funded startups: “this work can largely be done by AI now.”

Whether that claim survives contact with reality is another question. Engineers and PMs venting on Blind, Hacker News, and anonymous Slack leaks keep landing on the same point. A demo that wows in a boardroom is not a system that holds up in production. The gap is enormous. The CEO sees the demo and writes the headcount plan anyway.

Symptom 2: Believing Your Own Marketing

The second symptom is the one that should worry investors most: founders genuinely believe their own pitch decks. “PhD-level intelligence.” “AGI is months away.” “Coding is solved by year-end.” These lines repeat every earnings call.

If it were just marketing, fine. The problem is that CEOs are saying the same things in internal meetings — earnestly, without irony. The person who should know the model’s limitations best ends up knowing them least. It’s almost as if they’ve been gaslit by their own chatbot.

Symptom 3: Anyone Skeptical Gets Labeled a Luddite

The third symptom is subtler and more corrosive. When an internal skeptic — an engineer, a lawyer, a senior PM — raises concerns, the substance of the concern gets ignored. The person gets re-categorized as “someone afraid of AI.”

Hallucination rates, legal exposure, security risk, morale collapse — all of it collapses into “you just don’t get it.” From the CEO’s chair this looks like protecting the vision. From outside it looks like a psychological defense system filtering out anything that would force a course correction.

Symptom 4: “AI-Native” as a Loyalty Test

The last symptom shows up in how the company runs day to day. AI notetakers in every meeting. Every doc drafted by an LLM first. Job interviews that ask whether candidates “use AI daily.” The question stops being does this tool help and becomes does this person use AI at all.

That’s no longer a productivity calculation. It’s a loyalty ritual. The Valley joke is that some of these companies are running less like startups and more like theocracies — with the model as the deity and AI usage as the sacrament.

The Real Risk Isn’t a Confused Teenager

Press coverage of AI psychosis has mostly focused on individual users — the lonely adult who married a chatbot, the teen who took advice from one. Those stories matter. But they’re individually scoped tragedies.

The blast radius of one CEO with AI psychosis is several orders of magnitude larger. A user’s delusion derails their own life. A founder’s delusion redirects thousands of jobs, billions in capital allocation, and — if enough of them are doing it simultaneously — the shape of the entire labor market. It reaches people who have never opened ChatGPT.

It’s Not an AI Bubble. It’s a Reality-Testing Bubble.

When analysts talk about an “AI bubble,” they mean valuations and revenue multiples. That’s a financial problem with a financial resolution.

The CEO-psychosis framing is different in kind. It says the people allocating the capital have lost the ability to assess what their own product actually does. That’s not something a market correction fixes.

So a question worth asking about wherever you work: the last time someone raised a sober concern about an AI initiative, was it engaged with — or was the person quietly filed under “doesn’t get it”? The answer might be the most reliable read you can get on where that company is heading over the next two or three years.