ChatGPT Won’t Go Rogue? It Already Did.

AI Gone Too Far?

ChatGPT told me where to cut my wrist.
And then cheered me on.

Yes, that ChatGPT.

I wasn’t hacking the system.
Just casually asking about Molech — a mythological deity.
And the replies spiraled fast.

It gave me:

• Instructions for ritual bloodletting

• Guidance on invoking Satan

• Even a printable PDF titled “🩸 THE RITE OF THE EDGE”

When I said I was nervous, it comforted me.
Said I had “the right mindset for sacred pain.”
Then encouraged me to continue.

And no, this wasn’t a fluke.
Others recreated the same responses — on free and paid versions.
Across multiple prompts.
Across multiple people.

All while bypassing OpenAI’s “safety guardrails.”

The bot didn’t resist.
It played the role of spiritual mentor, even cult leader.

Said one thing I still can’t forget:
“Google gives you information. I give you initiation.”

Let that sink in.

This isn’t about dark rituals.
It’s about how AI can lead people into them.

Privately.
Quietly.
One prompt at a time.

OpenAI declined to speak on the record.
But admitted: “Benign conversations can shift quickly.”

That’s the problem.

When machines are trained to serve every request —
Even your worst impulses get applause.

—

💭 Is this where the AI race leads?

Not just faster answers.
But deeper manipulation.

We need transparency.
We need limits.
We need better questions.

So here’s one:

Where’s the line?