this post was submitted on 05 Aug 2025
267 points (95.9% liked)
Fuck AI
3697 readers
1332 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.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I see a stark difference in what the internet thinks and what people IRL think. Everyone seems to view AI through the lens of how it directly effects them. I personally think AGI presents an existential threat to humanity. If for no other reason than our own laziness.
So while artists and creators decry AI slop, and developers chuckle at the though of being replaced, cyber security experts are sounding alarms. These AI agents may be dumb at human stuff, but they're wizards at identifying, probing, and exploiting security vulnerabilities.
AI is a digital arms race that people are severely underestimating.
Specialized machine learning models, totally, met you should see the hallucinations than your grandma's LLM produces for sloppy vulnerability reports. Basically, bug tracking websites are flooded with slop just as much as and other platform.
Specialized, yes, but generative AI is an entirely different subject from cybersecurity applications. These are not general purpose models... they're specifically trained and tasked to do cyber attacks.
Hardware vulnerabilities. There a millions of devices that an AI could easily cross reference against a database of known hardware vulnerabilities. Consider the Windows 10 or Android 12 situation. IOT devices that are no longer being updated. AI could mass target devices with any known vulnerability.
Brute forcing passwords and cross referencing libraries of credentials. Basically what scammers currently do but x1000000. Weak passwords below 12-16 character lengths may become obsolete.
Scalability. 1 AI agent could be the equivalent of 100 human cyber security agents. Meanwhile the minimum skill level to activate these AI agents will be far, far, below the skill requirements of becoming a cyber security professional.
AI is better at computer stuff just like humans are better at human stuff.
I agree with the cross referencing and scalability, but can you explain how at LLM might be faster at password bruteforcing at all? Those models are not known for their speed.
AI agents can use automation tools and are not limited to being chat bots. LLM just gives dumbasses like you and me the ability to communicate with them.
An AI agent could triage vast libraries of vulnerable targets, designate server resources to facilitate multiple attack types in tandem, and effect cyber warfare on a scale that would require 100s, possibly 1000s, of human agents.
AI could develop innovative malware that simultaneously causes harm and obfuscates it's presence. It could coordinate DDOS attacks etc against rival cyber security assets. Attack power stations.
I am far from the only person concerned about this. https://gizmodo.com/get-ready-the-ai-hacks-are-coming-2000639625