Input crafted to fool a model into misclassification or unsafe behavior without obvious changes to humans. ← Account Abstraction Aztec Network (ZK Layer 2) →
