# Here’s how you know all these “AI Attacks” are fake AF — retained record

> Retained reddit record from reddit:OpenAI.

Canonical: https://highsignal.app/data/records/e4429be100992ec8

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:OpenAI)  published 2026-09-21 17:40 ingested 2026-09-22 02:37

[original source ↗](https://www.reddit.com/r/OpenAI/comments/1wmjqez/heres_how_you_know_all_these_ai_attacks_are_fake/)

## retained text



```
Notice how NOTHING EVER ACTUALLY GOES WRONG??

The model “hacks” Hugging Face but no data gets deleted. No money gets stolen. No one gets hurt. Just a perfectly safe “attack” that gets OpenAI a billion headlines about how powerful and scary their model is. lmao.

If they were actually scared of the model theyd run it on a computer with no internet. **It’s so fucking easy to contain a model guys. DONT PUT IT ON A MACHINE CONNECTED TO THE INTERNET.** 

Literally unplug the fucking internet. Boom. 100% contained.

Instead they give it internet access, put something “hackable” in front of it, **restrict what METR can audit**, then release the scariest possible version of the story.

Anthropic figured out that panic is AMAZING marketing. Every scary story makes Claude sound more powerful. OpenAI saw that shit, copied the strategy and one-upped them.

And the secrecy is the biggest tell. If the evidence actually proved their story, **why tightly control what independent auditors can see?** Why give us a curated narrative instead of the full logs and methodology?

Because they orchestrated the whole thing for publicity. DUH!

Real attacks destroy data, steal money and cause irreversible harm. These “attacks” somehow create maximum panic, maximum publicity and ZERO actual consequences every single time.

**This is not a “conspiracy theory,” it’s Occam’s fucking razor.** 

The gullibility people show in regard to these “attacks” is truly astonishing. 
```



document fetched 2026-09-22 00:27 · canonical https://reddit.com/r/OpenAI/comments/1wmjqez/heres_how_you_know_all_these_ai_attacks_are_fake
