Dumb But Secure · by alkizz
AI security for people who'd rather not.
Skeptical, practitioner-first security for AI as it's actually used: the agent on your spare box, the copilot your company just turned on, and the board asking what it all means.
be dumb to keep it simple
The pitch
Most AI security content is either a 90-page framework or a forecast of the robot uprising. This is neither. It's the boring stuff that actually works: scoped keys, closed ports, sandboxes, and calling people back before you wire the money.
Written for builders running agents and home labs, and for the businesses and CISOs who have to make it safe at scale. Dry humor included. Doom not.
Featured · For builders
The Builder's Checklist
For people running agents and home labs: keys, sandboxes, exposed ports, prompt injection, MCP supply chain, and a kill switch. Each item has a why and a how.
Read the builder's checklist →Featured · For businesses & CISOs
The Business Checklist
For companies and CISOs: shadow AI inventory, copilot permissions, vendor reviews, agent identities, and deepfake-proof processes. None of it needs a new committee.
Read the business checklist →Latest briefs
Example placeholders · not real posts- Builders
[Placeholder] Is your local model server actually local?
- MCP
[Placeholder] Pinning MCP servers without losing your weekend
- CISOs
[Placeholder] The one-page AI acceptable-use policy
- Agents
[Placeholder] Breaking the lethal trifecta, one leg at a time