Security thinking from the Liarbird team.
We took systems built for humans and asked AI to brute-force its way through them. It can. But the cost is staggering and the planet is paying for it. The fix isn't better AI. It's better architecture.
AI agents make attackers faster, more systematic and more thorough. They also make them more predictable, more exploitable and more fragile.
A North Korean intelligence operation compromised the most popular HTTP library on npm. The community caught it in three hours. What if they hadn't?
A state-sponsored group is embedding itself in telco infrastructure worldwide. Once they're in, they're nearly invisible. But getting there takes four noisy steps, and that's where defenders can still win.
The DoD called it active cyber defence. It means finding and disrupting attackers inside your own environment. The concept is older than you'd think.
The Trivy supply chain compromise turned a trusted security tool into an attack vector. No firewall was going to stop that. It's worth asking what your defences are actually pointed at.