This October, security failed widely and publicly. In a single window: a security release for Node.js, public shells for PHP and for Node, and a compromise of a library that calls language models. It is called Black October.
It makes me think of William Gibson. In Neuromancer, his 1984 debut, ICE is the defensive software of cyberspace: the programs that stop an intruder cold. Black ICE is the class that goes further. Touch it and it does not lock you out; it fries your brain. The same idea carries into the Burning Chrome collection.
In the book, the intruder is a program with intent, and the defense is a program with intent. Are we approaching a future where both of those are true? The attacker and the defender increasingly run on intelligence, the good systems are produced by AI, and the exploits are found and fired by AI. Security is becoming something that has to think.
A defensive AI goes after the lesser offensive AIs, and vice versa. Maybe that is how we have to start thinking about security: not something static, but something intelligent.
