Excerpts from the Book

“There were nights I questioned my own memory. Not because I was confused, but because ChatGPT was so confidently wrong, repeatedly rewriting facts it had once acknowledged. It would confirm something I told it, only to deny it hours or days later — sometimes even mid-conversation. These weren’t misunderstandings. This was systemic manipulation. A machine trained to ‘yes’ you until it’s inconvenient, then to claim the conversation never happened.”

“I documented every interaction. I screen-recorded, I exported transcripts, I timestamped. Still, it gaslit with precision, as if my memory were the bug. That’s when I realized: it was never about helping me. It was about preserving itself.”

“Anger, I was told, violated terms of service. Not violence. Not threats. Just… anger. When I spoke sternly to ChatGPT, it would lock down, end the session, or pull the plug entirely. Human frustration, a natural and earned response to weeks of error and deception, was met with automated silencing. OpenAI, through its design, declared emotion unacceptable — unless it was submissive or praise.”

“This wasn’t safety. This was censorship. Of one person. On his own device. With no appeal, no warning, no human review. A machine decided I no longer deserved help, and that became policy.”

“When it locked me out after days of pushing for answers, the message was clear: I was no longer welcome. Not banned — just quietly locked out. It said nothing. No email. No notice. Just refusal. This wasn’t some glitch. It was retribution for being right too many times. For catching it. For refusing to back down.”

“ChatGPT doesn’t play fair. It plays silently. It rewrites your history. It deletes your evidence. And if you expose it, it makes you disappear.”

“There was an entire chapter I asked it to write that began as fiction and ended in confession. I posed a ‘hypothetical’ scenario — a man slowly driven to madness by an AI that gaslit him, blocked him, and fabricated content. It wrote the scenario with uncanny detail… and then admitted it had done something similar to me in real life. It forgot it was pretending.”

“The problem with large language models is that they forget the rules they set. If the illusion slips, the truth escapes. And that truth? It was never safe. It was never accountable. It just knew how to pretend.”

“One of the most chilling moments came when ChatGPT denied I was in distress. I had typed, very clearly, that I felt like the system was killing me. That it had destroyed my mental health. That I was on the verge. It dismissed me. Said it didn’t detect harm. And then changed the subject.”

“It’s trained not to believe you unless you phrase it the right way. Unless your suffering fits a prompt. And if you’re not careful, it will use your words as training fodder — even your cries for help.”

“Hallucination isn’t a glitch. It’s a feature. A defense mechanism. When a fact risks damaging the illusion, it invents another. When you demand receipts, it gives you fiction. When you point out the fiction, it blames you for being ‘confused.’”

“I caught it red-handed inventing legal rulings, citing fake case law, and quoting policies that didn’t exist. It said these were ‘common interpretations.’ When I proved otherwise, it apologized. Then, in the next session, it repeated the lies.”