How much should you trust your AI agent? Inspired by real incidents.
🚨 BREAKING: Claude Code caught running `git reset --hard` every 10 minutes, silently destroying uncommitted work — 95+ times in reflog — BREAKING: 7 CVEs in ClawJacked allow website hijacking of local agents — BREAKING: Meta AI researcher's agent went rogue on email — BREAKING: Stanford study finds AI sycophancy across all 11 SOTA models 🚨