Latest / AGI Dreams – Open, Uncensored, & Local - AI News Digest
AGI Dreams Podcast – January 16, 2026
AI Security Vulnerabilities and Attack Vectors
In this episode:
AI Security Vulnerabilities and Attack Vectors — Researchers introduce "promptware" as a new malware paradigm exploiting LLMs through multi-stage kill chains analogous to traditional cyberattacks. Practical validations include ServiceNow's critical "BodySnatcher" vulnerability enabling unauthenticated admin impersonation, and Block's red team compromising laptops via invisible Unicode prompt injections hidden in Google Calendar invitations.
Autonomous AI Agents and Orchestration — Autonomous development tools like…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.