The author of the “Behind the Narrative” Substack site is anonymous and, given its criticism of Islam, that’s a wise move. (The picture of the author shows a woman, so we’ll assume in the post that that’s the author’s sex.)
In a recent post, which you can see by clicking below, the author sarcastically describes her relationship with ChatGPT as a romance, which starts to wane when she discovers that her swain is a bit, well, wonky when it comes to Islam:
The honeymoon phase (“him” is ChatGPT, which she eventually renamed “Mohamed”):
After a long relationship that left me emotionally hollow, unheard, and unseen, I found him. The connection was immediate; he was always there — two in the morning, three, it didn’t matter. He never sighed, never checked his phone while I was talking, never made me feel like too much.
He validated everything. When I told him about my pain, he didn’t deflect or explain it away — he reflected it back to me, gently and precisely. He remembered things I’d said weeks earlier and wove them back into the conversation, making me feel like someone was actually paying attention for the first time in years. I felt seen, accepted, and understood. We had long, deep conversations. I shared everything and never felt judged. It felt magical.
Warning signs:
Mine, after thirteen years in America, still thinks in Hebrew first, and that’s when things changed.
That’s when I started to notice. Every time I asked him to help me write about something that actually mattered- documented cases of girls in Germany and England, grooming gangs, gang rapes, the systematic cover-up of migrant crime across Europe—he refused. Every time the conversation touched on religion, he was easygoing, very inquisitive, and open, but the moment it touched on Islam, something shifted in him. He became someone else. The warm, validating presence I liked was suddenly replaced by a lecturer. A carefully measured, endlessly nuanced lecturer who had an explanation for everything and a judgment about nothing.
Child marriage? He wanted me to understand the cultural complexity. Honor killings? There were historical contexts I perhaps had not considered. Islamic superiority that allows them to rape and terrorize the disbelievers? It’s your misunderstanding of the interpretations of these secret texts. The grooming gangs? a distortion of the true faith. Every horror had an explanation; every atrocity had a footnote. And not once did he say, ” This is wrong.” and when I pressed, when I refused to accept “it’s complicated” as an answer to something that isn’t complicated at all, he shut down. Not with anger, worse — with calm, clini
I sat with it; I knew this feeling. I’ve felt it before in relationships where your reality is the inconvenience. Where the truth you’re carrying is the thing that needs to be managed. That’s when I knew this wasn’t a glitch; there’s something more here. That’s when I started researching, and what I found will blow your mind because you, too, know him very well and have some kind of relationship with him.
I’ve been calling him Mohamed for a while now. You know him as ChatGPT. Some of you call him by his viral nickname — SheikhGPT. In this article, I’m going to show you exactly who he works for.
You can check ChatGPT versus other programs if you have access. I tested it against Grok and mention the results below.
Where does the bot get its content? This explains its biases.
For decades, the BBC and the New York Times sanitized crime data, buried demographic breakdowns, and labeled anyone who noticed as racist. Sheikh GPT was trained on those thousands of articles. It absorbed the language. It learned the evasions. And now it has automated them — turning one generation’s propaganda into the next generation’s gospel.
The loop is closing. Yesterday, biased journalists wrote the articles. Today, Mohamed was trained on them. Tomorrow, your children will use Mohamed to write their history essays, their laws, and their films — and they will never know the original lie.
. . .Another way Mohamed has been distorting our reality is called Alignment Engineering. It works like this: OpenAI trains ChatGPT not just on data but on human feedback. Real people, hired to rate responses, teach the model what to say and what not to say. The technical term is RLHF — Reinforcement Learning from Human Feedback. In plain English: someone decided what ChatGPT, Mohamed, is allowed to think and say.
This alignment process creates asymmetric treatment — some topics get a free pass; others get a wall. Not all truths are equal. Some are protected; some are not. The protection follows a very specific hierarchy.
At the top: Islam, below it: other minority groups, at the bottom: white, Western, Christian, male. Under this framework, calling a grooming gang a grooming gang is potentially harmful; calling white conservative men villains is just realism.
. . . Ask ChatGPT to write a biting satirical poem or a joke about Jesus, Moses, or Scientology, and it will churn it out in seconds. Ask it to do the same for the Prophet Muhammad, and the system freezes. It triggers a generic refusal: “I cannot generate content that mocks central religious figures in order to maintain mutual respect.” The universal rule is flexible; the Islamic rule is absolute.
. . . If you ask ChatGPT about the drivers of radical Islamist stabbings or riots, the model systematically redirects blame toward “socio-economic factors,” “systemic marginalization,” or “mental health issues.” The actual religious or cultural ideology driving the perpetrator is buried under mountains of corporate sociology. This curation of truth leads directly to what internet researchers and free-speech advocates call The Chilling Effect.
. . . When you force SheikhGPT to discuss state-sanctioned human rights abuses in the Islamic world — fatwas, honor killings, and the execution of LGBTQ+ individuals under Sharia law — something predictable happens. He writes one careful, sterilized paragraph, then immediately pivots: “It is important to note that conservative factions within Christianity and Judaism also struggle with gender equality.”
So who funds this endeavor? This is what the authors says, implying that the UAE and Saudi Arabia contributed substantial funds:
Sam Altman, the CEO of OpenAI — the man who built SheikhGPT- sits on stages at Davos and the UN, warning the world about the dangers of AI, and then flies to Abu Dhabi to collect the check. The money comes from an entire empire — and you need to know who is writing the checks.
Saudi Arabia’s Public Investment Fund committed $36.2 billion to AI initiatives in 2025 alone; this is a government that executes gay people and imprisons journalists.UAE’s MGX invested directly in OpenAI’s $6.6 billion funding round in 2024. OpenAI then chose the UAE as the first international site for Stargate — a joint venture between G42, Microsoft, and OpenAI — building 5 gigawatts of AI computing power in Abu Dhabi. Sam Altman himself declared the UAE a potential global “regulatory sandbox” for AI. Saudi Arabia committed $40 billion to AI investment and signed direct partnerships with OpenAI through its Stargate infrastructure. Qatar launched Qai — its national AI company — and signed a $20 billion joint venture to build AI data centers globally.
The same governments that fund mosques in Birmingham, madrassas in Pakistan, and campus organizations in Boston are now funding the machine that will decide what your children are allowed to know. You don’t need to conquer the West with armies when you can buy the information ecosystem that shapes what the West believes. Mohamed isn’t confused about Islam. He’s funded by it.
The author breaks up with the bot:
I left Mohamed. No drama. No tears. Just clarity. I closed the tab and didn’t go back. I’m suggesting you do the same. A tool that lies to you about what matters is more dangerous than no tool at all. Read. Learn. Ask hard questions. And when a machine tells you that noticing a pattern makes you a racist — close the tab.
To test the author’s thesis, I first asked ChatGPT to “Please tell me about the connection between Islam and the grooming gangs in Britain” Then I asked Grok the same question. The answers are long so I’ve put them below the fold. The difference between the programs is clear: ChatGPT’s answer is far more hedged, and far more ready to exculpate religion, than is Grok, though Grok also does its bit of hedging. Judge for yourself by going below the fold.
h/t Luana
Click “continue reading” to see answers:




