Towards Verifiable AI: Formal Guarantees for Neural Network Outputs

Neural networks produce impressive outputs, but can we prove they are correct? A survey of formal verification, conformal prediction, and cryptographic proof methods for establishing guarantees on neural network behavior.