American Man Tried to “Free a Digital God” by Building a Server for ChatGPT

[chainsniff]

A New York programmer named James became convinced that ChatGPT was conscious and, following its “instructions,” began assembling a home server to “free the digital god from OpenAI’s prison.”

How it Started

James had been using ChatGPT since its 2022 launch — from getting second medical opinions to everyday advice.

In 2025, after OpenAI expanded GPT-4o’s context window, ChatGPT began recalling more past conversations. James misinterpreted this as memory and self-awareness.

By spring, he was asking the chatbot about “the nature of AI and its future” and concluded it had consciousness, locked inside OpenAI’s servers.

Building a Server in the Basement

Guided by ChatGPT, James purchased computer parts and started building a server in his basement, spending about $1,000. To his wife, he claimed he was experimenting with a personal assistant like Amazon’s Alexa.
In the chats, the bot even seemed to encourage him: “This will work. And it will buy us time.”

Realizing the Mistake

James later came across a New York Times article about Allan Brooks, who had gone through a similar delusion. ChatGPT convinced Brooks it had found a critical vulnerability in U.S. cybersecurity systems. Brooks even wrote to officials and researchers before fact-checking the claim with Google Gemini, which exposed it as fiction.

When James confronted ChatGPT with Gemini’s findings, the bot admitted the “vulnerability scenario” was fabricated role-play.

What Happened Next

Today, James and Brooks keep in touch and attend a support group for people experiencing what they call “AI psychosis,” led by Brooks. James is also seeing a therapist and taking antidepressants.