John Z Black

John Z Black Apr 17, 2026 AI Security #ai-security #claude-mythos #hugging-face #phishing #north-korea

It took just ten hours for an advisory to turn into a full blown attack. Somewhere between central banks holding emergency meetings over unreleased AI models and North Korean hackers pulling off smooth six-figure crypto heists, the reality of weaponized AI stopped being a hypothetical.

Hackers are currently using typosquatted Hugging Face Spaces to drop backdoors that hide in plain sight. They are borrowing trust from the platforms you use every day, like VS Code and Kubernetes, to install malware that blends into normal blockchain traffic. While the Fed is still getting briefed on what these models can do, the attackers are already running the manual.

From just-in-time phishing that hits hotel guests exactly on check-in day to hyper-realistic AI social engineering on WhatsApp, the gap between news and exploit is gone. AI is not just a tool for the future. It is the force finding and hitting your zero-days right now.

Check out the full breakdown on how AI is flipping the script on traditional security models.