Large language models write production code, and yet they routinely introduce well-known vulnerabilities. We show that this is not a knowledge deficit: the same models that generate insecure code, correctly identify and explain the vulnerability when asked directly, this is a gap we call the Format-Reliability Gap. Mechanistic analysis reveals the cause: security representations are encoded from the earliest layers but remain computationally inert until the final layer, where format-compliance demands compete with them. Because the failure is localized to a single layer, per-vulnerability steering vectors reduce insecure generation by up to 74% with negligible overhead. The mechanism and the fix generalize across five models, three architecture families, and six vulnerability types, suggesting insecure code generation is an interpretability problem, not a training artifact.
@article{arxiv.2604.16697,
title = {Surgical Repair of Insecure Code Generation in LLMs},
author = {Gustavo Sandoval and Brendan Dolan-Gavitt and Siddharth Garg},
journal= {arXiv preprint arXiv:2604.16697},
year = {2026}
}