Impart uses AI to build virtual patches without putting models in the traffic path

Impart Security has introduced a new product called the Runtime Defense Harness, designed to let AI agents draft and test virtual patches for freshly disclosed vulnerabilities without ever placing the AI model directly in live traffic. In this setup, agents analyze real customer traffic to build a candidate patch, a human engineer signs off on it, and Impart's enforcement engine then runs it as deterministic WebAssembly code, delivering a tested fix in as little as 60 minutes rather than the days or weeks typical of vendor signatures or code-level fixes. The company points to the recent PHP2Shell vulnerabilities as an example of urgency, noting attackers began exploiting them four days before public disclosure while roughly 80% of 4.38 million internet-facing PHP services remain on unsupported versions. Impart also cites a high-severity Node.js flaw, CVE-2026-111497, where its customers received a blocking patch within 24 hours, compared to a major CDN that took four days and shipped only a log-only rule.