Notes from the runtime
Writing on AI security from the team building the runtime: engineering, governance, and the road to compliance you can prove.
First Recon AI launches AI Security Runtime
The company announces its public launch and the availability of AI Security Runtime. Four functions run on every AI interaction, from any device to any model, and each enforcement decision is recorded as evidence.
Read the announcement July 2026 · PerspectiveThe unit of AI security is the interaction
Files, networks, and endpoints each got a security discipline of their own. AI interactions are the new unit of work, and they need one too.
Read the essay June 2026 · EngineeringAnatomy of a prompt injection
Direct injection, indirect injection, and tool-call hijacking, stage by stage, and where inline enforcement breaks the chain.
Read the breakdown June 2026 · ProductShadow AI is a preference problem
Unsanctioned AI spreads because the tools are good. Field notes on why blocking loses, and what a governed workspace has to get right.
Read the field notes May 2026 · ComplianceWhat audit-ready should mean for AI
A policy document says what should happen; a decision record shows what did. On sealed evidence and answering auditors from the trail.
Read the essay