# AI Immune Systems
AI Immune Systems is the assurance layer of [[Cognitive Sovereignty Infrastructure]] that treats machine cognition as a contested surface rather than a trusted tool.
## Threats addressed
Data poisoning, adversarial examples, model stealing, jailbreaks, corrupted recommendation, compromised assistants, electronic-warfare effects, and silent degradation of decision-support systems.
## Program lineage
[[DARPA GARD]] developed robustness tooling; [[DARPA SABER]] applies recurring operational AI red-teaming; [[AI Cyber Challenge]] demonstrated autonomous vulnerability discovery and repair; [[IARPA ReSCIND]] turns attacker decision biases into a defensive surface.
## Operating principle
Assurance is continuous: test, observe, challenge, recover, and retest. A benchmark pass or vendor assertion is not durable evidence that a deployed model remains trustworthy.
## Source
[[articles/Cognitive-Cyber Warfare Measures and Countermeasures#AI Immune Systems|Cognitive-Cyber Warfare — AI Immune Systems]].