Last released Oct 11, 2026
AI security reviewer that proves the fix: exploit it, patch it, verify it — on an open model, offline.