Read our audit of AI-generated memory-safety proofs, which examines the guarantees behind Anthropic and Theorem’s verification results.