Who runs the doomsday machine?

Calling AI a doomsday risk fits well with calls to centralize and limit access to it. view full post

AI comapnies have leaned into the doomsday messaging.

They point to the hugging face hack1 ( my response ) as evidence that AI could “take over the internet”.2 But how?

“it’s my worry that in 6-12 months such a swarm could be capable of taking over the entire internet with a persistent botnet”.

– Dario, who built the program to do this and runs it all day every day

Those companies run the AI that would take over the internet. They run the computer program they are afraid of. All day every day. They build it to be capable of taking over the internet and run it on their machines. If it did take over the internet, they could shut it down. They can watch the logs of what it is doing. They can read the IP addresess it connects to. They themselves can limit internet access whenever they want.

They also release reports about people trying to use their AI to do bad things like build missiles3.

We do not have evidence the actors succeeded in fielding an operational device; but they did test-fire a guided rocket. This field test appears to have failed: within hours, the actors returned to Claude to work out why it failed.

Again, they know this is happening because they control the AI.

They call for “regulation” and “slowdown”, but they could just shut down the whole thing. They want us to be afraid of AI, but they want to continue to run AI for us. They want to regulate AI, but right now as they are already making tons of money and are about to IPO.

Why!?

They want to regulate away competition. They’d like to ban open models - those are truly terrifying, right? They can run on your machine, not the cloud. No one watches over them but you. No one can ban you from them. They are developed in labs around the world, not here under the watchful eye of US government investigators (conveniently from AI companies).

First they make you afraid that you dont use enough AI - you’ll be left behind man!

Then they make you afraid of people that don’t use their AI - you know, the one that’s safe man!

It’s a bizzarro second ammendment debate all over again for the 21st century.


  1. https://openai.com/index/hugging-face-model-evaluation-security-incident/ ↩︎

  2. https://darioamodei.com/post/we-must-pace-the-frontier ↩︎

  3. https://www.anthropic.com/threat-intelligence-report-september-2026 ↩︎

Comments

I have not configured comments for this site yet as there doesn't seem to be any good, free solutions. Please feel free to email, or reach out on social media if you have any thoughts or questions. I'd love to hear from you!