PentAGI just launched an open-source AI penetration testing tool.

Anthropic just launched Claude Code Security.

AI is entering cybersecurity — permanently.

But here’s the uncomfortable question:

What happens when attackers automate faster than defenders react?

Maybe the answer isn’t faster response, but removing default visibility altogether.

Some time around 2010 I update th virus and Trojan definitions on a windows 95 machine that wasn't really used beyond 2004. It had every free scanner one could download. They found over 1500 things. Apparently someone ran a trojan generator on my machine. I don't know when they all became known trojans or if it found all of them but when the box was in use the scanners found nothing.

God knows what is sleeping on our machines today.

I once hear there are very very expensive tools with luxurious gui's that bundle bags of fresh exploits and receive new ones over the tubes.

The guy told me that most of the code involved cleaning up after it self.

I think the AI will only run known exploits? Its nice to find those but if anyone really wants in it's just a matter of money.

I'm curious what applications we can bake into hardware. You can't insert malicious bread into a toaster to gain access to other things.

Apparently the formal specification of OpenNHP was submitted as an IETF Internet-Draft in January:

https://datatracker.ietf.org/doc/draft-opennhp-saag-nhp/

Not quite. Cixin Liu’s series had a premise of dramatic asymmetry.

AI in our context is the inverse. Everyone can spend “credits” to get a supergenius coder.