Anthropic Says Claude Proved Security Properties Across 123 Libraries

Anthropic announced that its AI model Claude generated formal machine-checked proofs verifying security properties for 239 fuzz-test entry points across 123 open-source libraries at the RISC-V assembly level. Claude also patched bugs and proved the programs exit normally without memory-safety errors under specified conditions, marking a first large-scale AI application of formal verification at assembly code level. Anthropic published a subset of these proofs and related components on GitHub.