AI Watchtower

AI Watchtower is a critical-watch project on the security of AI systems. Each entry takes a piece of ongoing research — direct and indirect prompt injection, MCP server vulnerabilities, agent threat models, concrete hardening tradeoffs — and turns it into actionable guidance for builders and operators of LLM-based assistants.

The format is short, structured, operational: one analysis, one risk grid, one checklist you can apply the same day.

Articles below.

Claude Desktop — Simple Hardening: What Claude Should NOT Have Access To

Targeted attack-surface reduction for Claude Desktop. Simple principle: it’s a chat assistant, not a system agent. The list of NOs, the list of OKs, and a 30-minute checklist.

April 27, 2026 · 10 min · 1935 words · aleph-beth