Prompt Injection Attacks Are Thwarting AI Hacking Agents
Wired
—
“Context bombing” tricks malicious AI agents into shutting down before they can do harm.