Proving the fuzzing tests are crash-free seems like a nice idea -- significantly easier than cooking up high-level specs
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.View quoted post on X
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: anthropic.com/research/forma…