AI-generated Rust compiles perfectly. That’s the scary part.

Canonical is investigating whether automated tools can rewrite legacy C code into safe and maintainable Rust without altering the software's functionality. Researchers from the University of Bristol are testing this with AppArmor and snap-confine, which provide security-focused features and sandboxed environments. The ultimate goal is to generate Rust code using language models, then verify its behavioral equivalence to the original C code.

The Bristol team will combine fuzzing with formal program analysis to detect and repair differences between the generated Rust code and the original C code. Canonical is funding this research to determine if the approach can be scaled to large repositories of C code. Trust is the key hurdle, not the code generation itself. The researchers will also ensure that the translated Rust code maintains the same level of security as the original C code.