GuardBreaker uses safety-sensitive text inside malicious scripts to distract AI security tools from malware. The technique ...