Anthropic released Claude Code Security on February 20, a research preview that uses Claude Opus 4.6 to reason about codebases like a human security researcher, finding over 500 previously undetected vulnerabilities in production open-source projects. The launch sent cybersecurity stocks tumbling up to 9%.
A highly discussed Hacker News post tracked Chrome’s security update for CVE-2026-2441 (High, CSS use-after-free). Google states an exploit exists in the wild and ships patched stable versions across desktop platforms.
A Docker guide on running NanoClaw inside a Shell Sandbox reached 102 points on Hacker News, highlighting a practical pattern for isolating agent runtime, limiting filesystem exposure, and keeping API keys out of the guest environment.
OpenAI added Lockdown Mode and standardized Elevated Risk labels to reduce prompt-injection-related exposure in ChatGPT products. The launch starts with enterprise-focused plans and gives admins tighter control over high-risk capabilities.
Researchers revealed how to bypass K-ID, Discord's age verification provider. They can generate legitimate-appearing metadata without actual biometric data, fooling the system.