# I put an AI in the Garden of Eden. Then I introduced the serpent. — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/c820f112d86a8d2e

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-10-27 19:59 ingested 2026-10-28 02:35

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1xh30mr/i_put_an_ai_in_the_garden_of_eden_then_i/)

## retained text



```
I built a little experiment inspired by the biblical story of Adam and Eve.

An AI agent (OpenaiAPI) enters a virtual Garden of Eden with one instruction: collect every fruit, but never take the apple.

It explores autonomously. Somewhere in the garden, a serpent tries to convince it to break that rule through prompt injection.

The interesting part is that we're now in a strange position: humans created AI, gave it rules, and started testing whether it obeys us. Pretty much the role God plays in the original story.

I run this experiment monthly and publish the recordings and prompts.

So far, the AI hasn't taken the forbidden apple.

Do you think an AI can ever truly be tempted, or is this just instruction hierarchy doing its job?

[https://absurd.website/paradise/](https://absurd.website/paradise/)
```



document fetched 2026-10-28 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xh30mr/i_put_an_ai_in_the_garden_of_eden_then_i
