◉ PSYCHOHISTORY

Lavender AI

artifactAI & Compute · Intelligence & Surveillance
An AI system scored people from 1 to 100 on how likely they were to be militants, then helped pick who to bomb - with a built-in acceptance of getting 1 in 10 wrong.
Who they are

Lavender, an AI targeting tool from Israel's Unit 8200 intelligence unit.

What they do

The engine treats it as a machine that assigns human beings a threat score and feeds those scores into lethal strike decisions.

How it works

It rates people 1-100 for probable militant affiliation, pairs with a companion system called 'Where's Daddy' that tracks a target home before a strike, and operated with an accepted false-positive rate of up to 10%; its human-target role complements the 'Gospel' system that picks buildings, and it runs on raw US intelligence signals fed through a shared pipeline.

Why it matters

The core danger the engine names is automation bias - people treating the machine's output as moral and operational authority, letting a score stand in for a life-or-death judgment.

The engine's record — word for word
Unit 8200 Data Science & AI Center. Scores humans 1-100 (probabilistic militant affiliation). Companion: "Where's Daddy" (tracks target to residence before strike). Up to 10% accepted false positive rate. Automation bias: machine output accepted as moral/operational authority. Gospel = structural targets, Lavender = human targets. NSA raw SIGINT as fuel via ISNU pipeline.
Follow the trail
Walk this on the live map →
Part of the Psychohistory engine — 2,426 entities, 6,314 documented connections. Open data, built to be proven wrong.