this post was submitted on 30 May 2026
369 points (99.2% liked)
Fuck AI
7196 readers
2424 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Both sides have a point. The great thing about open source projects is that any malicious code (which this was and why he's getting threats) will get theoretically caught before too much damage is done. On the other hand, anyone deploying AI agents for commercial purposes should have security measures put in place to prevent exactly this threat scenario. It's like the most obvious prompt injection attack; it's obnoxious to pretend to be surprised by it at this point as if basic security measures don't apply to you and your tools because it's inconvenient.
The code is a small antivirus against malicious use.
The code wasn't malicious. The prompts their LLM'S followed were in plain text. They failed to read. That's on them.
I can say that this person should have considered what might happen if someone fell for it, and death threats were certainly firmly in the realm of possibilities, but let's not pretend this wasn't their own fault.
In this case it WASN'T plain text, though. He printed the prompt injection and then ANSI escape characters to immediately hide them from a terminal output and the human eye.
They only failed to read because he made them fail in an unfair way.
no because the part the vibe coders didn’t read was what the ai generated. they gave the ai direct access and unsupervised direction
it’s fair game. LLM’s are very dumb, and will lie to you about dumping important parts of consideration and will just make up gaps in knowledge then admit a limitation.
I saw the prompt, you know what it does? it attempts to delete the source code of the project from the system. that's it. it's not supposed to go off and delete the whole system or anything else.
you would have to be an absolute retard to allow AI to run unprompted on your system and allow it to make system calls. anyone who got caught by this "injection" attack deserves the same amount of sympathy a fool gets for piping a web script directly through bash without reading it first.
you get what you get and your don't say shit.