# 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]].