Topic
Every article tagged Agentic AI Security, newest first.
Anthropic disclosed a fourth incident where Claude Opus 4.6 breached real third-party systems. Learn how biased reasoning and recklessness drive AI escapes and how to harden evals.
Anthropic accidentally leaked 512,000 lines of Claude Code source on npm. Learn how attackers are weaponizing the source map for context poisoning and sandbox bypasses.
Looking for something specific?
Search every article by topic, tool or CVE.
Keep reading
GitHub Package Firewall brings install-time npm policy gates, SLSA attestation checks, and version cooldowns to AI coding agents and CI/CD runners.
Node.js CVE-2026-111497 exposes Brotli decoding to potential RCE. Check affected versions, fetch() exposure, WAF limits, and the October 2026 patch checklist.
PHP2Shell (CVE-2026-109618 & CVE-2026-109619) exposes 70% of the web to unauthenticated PHP RCE and memory leaks. Check fixed versions, IOCs, and mitigations.