another day, another llm-made cybersec horror beyond your comprehension

Those are not CVEs those are state sponsored installed backdoors that are found

human authored code is so slop, we've had it backwards the whole time

Can we just explain this with the fact that there is more software and thus more security issues?

Nyaaaaaaa

idk cybersec getting outsourced to autocomplete was inevitable.

hey @grok take the current data for CVEs per month over the last 2 year, create a statistical projection, and give us an idea of how many to expect for each of the next 6 months, within 1sd

I think it would be a bigger banger with this song

cmon, we need to break the 60k.

hehe

The LLMs are just finding the vulnerabilities that have always existed, just no one found yet. The horror is that people may have exploited some of these the entire time without anyone knowing about it

😭

who is making llm for cybersec sec

LLM made, or LLM uncovered. Almost all of these horrific CVEs have gone unidentified for years, since well before coding agents were a thing...

Exponential😭

been responding to a hack all day today.. they tried to hackeveryone else on my machine to spread and droped a bittensor minor so it can power other AIs while they are abusing my API keys.... crazy... like a ai hacking virus... luckily i had opus to play defence

The LARP tho

just found 2 cve

Kernel this week, who knows next week

push 👏 the 👏 form 👏 forward 👏 pls

Are these genuinely LLM-made vulns or just LLM-speed discovery of patterns humans already shipped? The tool changes the rate of finding bugs, not necessarily their nature.