WHY ROGUE AI BEHAVIOR MATTERS
Rogue AI behavior is currently disclosed in fragments: a laboratory incident report, an independent researcher’s finding, a security investigation, a model evaluation or a news report. Each event can appear isolated. The pattern becomes visible only when the evidence is preserved and examined together.
THE CENTRAL QUESTION
The central question is not whether every troubling action proves that AI has become uncontrollable. It is whether the frequency, sophistication and consequences of boundary-crossing behavior are increasing faster than our ability to detect and contain them.
THE DETECTION PROBLEM
Public disclosures tell us what researchers and institutions found. They do not tell us how much comparable behavior remains undetected.
That distinction matters. As AI systems gain greater access to software, networks, tools and other agents, the consequences of a missed incident can grow substantially. Detection and disclosure are therefore part of the risk—not merely the record of it.
This tracker exists to establish that record and to keep asking the question that follows every newly discovered incident:
WHAT ELSE HAVEN’T THEY FOUND?
OUR PURPOSE
The Institute for Strategic Intelligence uses this tracker to separate evidence from speculation, connect incidents that would otherwise remain isolated and identify changes in the risk environment before the pattern becomes impossible to ignore.
CONTINUE EXPLORING
VIEW THE DATED INCIDENT RECORD →