Topics
Briefs and stories grouped by subject. Some stories fit more than one.
Jailbreaks, indirect prompt injection, leaked system prompts, and attacks that ride in with retrieved text.
Poisoned weights, malicious model files, and the packages and hubs those models travel through.
Tool use, agent hijacks, and the permissions an automated assistant gets on a machine or an account.
Laws, regulators, and public guidance that change how AI systems are built, bought, or disclosed.
AI Security Guru