Threat landscape
The AI threat landscape, annotated by people it's aimed at
Generative AI made scam copy fluent. Here's what changed in the messages we scan, and what still gives them away.
Blog
SIGNAL.
Dispatches from the engine room. Written by the people who read scam traffic for a living.
The engine
An engine that grades itself: self-learning defense, explained
Every miss becomes training data. How the engine reads live scam traffic, retrains on what slipped past, and tests itself before it judges your messages.
The registers
How scammers write: eight message families we train against
Ajo circles, dispatch riders, BVN threats, airdrop bait. Each legitimate family has a scam twin built to look identical. We make both, on purpose.
Infrastructure
We shipped post-quantum encryption before anyone asked
Hybrid X25519 + ML-KEM, the same NIST-standardized algorithms as Signal and iMessage, and why a scam-detection app bothered.
Case file
Anatomy of a wallet drainer: the “claim” link that worked
A lookalike domain, a countdown, a community of believers. A walkthrough of the web3 scam our engine caught in the wild.
The fortress
No side left open: what a digital fortress actually covers
Texts, chats, links, browser, inbox. Where each attack comes in, which guard stands there, and what still gets through when you rely on only one.
These are real drafts from our research files, publishing with the site. No filler posts, ever.