AI agents trigger internet-scale vulnerability cascade

AI coding agents can rapidly scan decades of legacy code, find critical zero‑days (example: PHP flaws), and either publish or weaponize them so quickly that patching and incident response cannot keep up, producing cascading outages or mass website takeovers. The speed, low cost, and global availability of open models converts latent technical debt into immediate systemic risk. — If true, this forces new public‑policy choices about model export, software stewardship, national patching protocols, and the governance of open‑weight AI, with direct consequences for infrastructure, elections, and commerce.