Can AI prove code is bug-free?

A new blog post from an unnamed author examines what formal verification of code can and cannot guarantee, arguing that the value of a proof depends on the specification it is asked to establish. The post points to a checked example and to findings from an audit of the Anthropic–Theorem verification effort, with the paper and supporting evidence published in the focs-lab/memory-safety-proof-audit GitHub repository. The work matters because it tests claims that AI-assisted verification can show code is bug-free against the memory-safety specifications actually proved.

Can AI prove code is bug-free? What a proof guarantees depends on what we ask it to prove. Our new article https://focs-lab.comp.nus.edu.sg/blog/memory-safety-audit/ explains why specifications matter, using a checked example and findings from our audit of the Anthropic–Theorem verification effort. Read the paper and supporting evidence https://github.com/focs-lab/memory-safety-proof-audit .