Episode Details

In this practitioner-focused briefing, we examine the shifting landscape of local AI security and autonomous threat actors. We break down the 'Bleeding Llama' vulnerability (CVE-2026-7482), a critical heap out-of-bounds read in the Ollama framework that exposes process memory. The discussion extends to unpatched remote code execution flaws in Ollama for Windows and a significant data exposure at Zara linked to the Anodot incident. Our primary analysis focuses on a new report from Google's Threat Intelligence Group, which details the first identified zero-day likely developed by AI and the use of agentic tools like Hextrike by North Korean and Chinese state actors.

Topics Covered

• 🔒 Bleeding Llama and memory safety risks in local LLM inference frameworks.
• ⚠️ Unpatched path traversal and signature verification flaws in Ollama for Windows.
• 🛡️ Google GTIG report on AI-developed exploits and autonomous attack orchestration.
• 🚨 Analysis of the Zara BigQuery data exposure and the Anodot supply chain link.
• 🌐 TrickMo Android malware's adoption of the TON blockchain for stealth C2.

The information provided in this podcast is for educational purposes only and does not constitute professional security advice.

Neural Newscast is AI-assisted, human reviewed. View our AI Transparency Policy at NeuralNewscast.com.