C-to-Rust tools failed to compile 342 of 464 translations and carried memory bugs into 177.

arXiv 2609.25682 ran C2Rust-analyze, CROWN, C2SaferRust and FLOURINE, including LLM approaches, on 116 NIST Juliet programs with known memory bugs. 77 outputs introduced new Rust bugs. The tools meet their own definition of memory safety, meaning less unsafe, and miss memory security entirely. MIT Lincoln Laboratory released DARPA's TRACTOR benchmark the same day, which at least gives this a shared yardstick.