Security AI

Security at Machine Speed Is the Wrong Race

Machine-speed exploitation will outrun reactive defense. Security invariants create resilience by eliminating attack paths before an incident begins.

AI Security

When AI Safety Becomes a Competitive Moat

Anthropic wants government to turn frontier AI into a permissioned market. Determined users can route around the gate through foreign and open-weight models, while compliant American companies bear its costs.

Security Hacking

Escape QEMU: Watching IronCurtain and an Open-Weight Model Break Out

A concrete run of defensive vulnerability research on open weights you control. Driving IronCurtain's workflow, GLM 5.2 builds a reproducible proof-of-concept escape of QEMU's public EDU teaching device from analysis to host code execution. The capability benefits from orchestration and does not depend on any one model.

Security Hacking

Two Talks: AI Zero-Days and Security Invariants

Slide decks from two recent talks built on IronCurtain, embedded for reading here.

AI Security

The Case For Open-Weight Models And Why We Can't Trust Frontier Labs

A frontier API can refuse, change, or vanish out from under you. Open weights keep the model you depend on yours.

Security Hacking

The Day After the Zero-Days

Patching faster cannot keep up with AI-driven discovery. Leverage structural invariants to make bug classes irrelevant.

Security

The Arrangement

A story that changes depending on who is reading it.