Tag
Chainguard has expanded its Agent Skills offering with a public registry of over 1,000 hardened AI agent skills, a private registry, and a hardening-as-a-service tier, aiming to bring secure-by-default practices to AI coding agents.
A security expert shares a cheatsheet on advanced agent security hardening, covering tool sandboxing, output validation, data loss prevention, adversarial testing, and runtime policy enforcement, emphasizing continuous security practices for production AI agents.
Sécurix is a NixOS-based secure operating system developed by the French digital department (DINUM) for hardened environments with strong isolation, reproducibility, and policy-driven configurations, currently in alpha.
C++26 is introducing standardized library hardening to catch common undefined behavior (like out-of-bounds access) at runtime, based on Google's production experience showing a mere 0.30% performance overhead and a 30% reduction in segmentation faults.