Prompt Injection: Making AI Do Things It Shouldn't
Direct and indirect prompt injection in LLM applications — real attack examples, vulnerable LangChain agent code, OWASP LLM01, MITRE ATLAS, detection, and …
Posts tagged: Ai-Security
CISA Mandates Patch for Windows Shell Zero-Day CVE-2026-32202 — GitHub RCE and LiteLLM SQLi Also in Play
CVE-2026-32202 Windows Shell zero-day confirmed exploited; CVE-2026-3854 GitHub RCE; CVE-2026-42208 LiteLLM SQLi hit in 36 hours.
Microsoft Defender Zero-Day Exploited Pre-Disclosure
CISA adds Microsoft Defender zero-day to KEV. Unit 42 Zealot AI agent pwns cloud. CanisterSprawl npm worm self-propagates. Talos Q1 2026 IR data.
Vercel Breached via Context.AI Supply Chain Attack
Vercel breach via Context.AI supply chain attack, Anthropic MCP protocol RCE flaw, QEMU emulator abused for ransomware evasion — April 20, 2026