llm-security

2 posts tagged "llm-security"

AI Needs the Context Auth Was Built to Limit

Access controls were built for two cases: automated workloads with a predictable footprint, or a human stitching tools together by hand. AI agents are neither, and the Vercel incident previews what breaks when we stretch the old model to cover them.

Ungoverned Context Is a Real Supply Chain Risk for Agentic Workflows

Ungoverned context is a real supply chain risk for agentic workflows. Most teams can identify the agent's runtime but have no way to reconstruct what sources actually shaped the output.

← All posts