First lessons live · more shipping
AI security, the version that holds up under real use.
Most AI security content is written for an audience that runs the model. Most users do not run the model. They use it, build with it, agent it, paste secrets into it, and inherit its mistakes. We are writing the lessons for the second group, and the first.
Live now
- 01Prompt injection 101: how attackers talk your AI into betraying you
Prompt injection is the SQL injection of the AI era. The attack is not on the model. The attack is on the trust you placed in the model. We unpack the shape so you can spot it before it costs you.
9 min
- 02How scammers use AI now
AI did not invent fraud. It made the old scams cheaper, faster, and far more convincing. Here is what actually changed, and what did not.
8 min
- 03Deepfakes and voice clones, explained
A thirty-second clip of someone speaking is now enough to fake their voice. Here is how the fakes work, where they fail, and the one defense that does not depend on your ears.
9 min
- 04Spotting fake support and fake profiles
The fastest way to lose self-custodied funds today is to ask for help in public and trust whoever answers. Here is how the fake-support trap works and how to stay out of it.
8 min
- 05How to verify anything in the AI era
When anyone can generate a convincing article, screenshot, or video in seconds, the skill that protects you is verification. Here is a practical method you can run in under a minute.
9 min
- 06Using AI tools safely as a creator or business
AI tools are genuinely useful, and they leak data in ways most people never notice. Here is how to get the upside without handing your work, your clients, or your keys to a third party.
10 min
Why this domain
AI is the next domain where the same self-custody question shows up: who holds the keys to your data, your prompts, your decisions, and your reputation. The threat model is new. The principles are not.
Where the pillar is going
- Prompt injection, plain English version
- Agent permissions and what they actually authorize
- Data exposure through chat, embeddings, and tool calls
- Model risk for builders shipping with LLM APIs
- Jailbreaks, social engineering of AI assistants
- Auditing what an AI assistant has done on your behalf
Early access
Get the first lesson the day it ships.
First lessons shipping next.