Sep 17, 2026 · 1h 13m
Humans turn AI into a cyber weapon
The biggest AI threat still comes from humans | E31
The episode reframes AI risk around the criminals, companies, and governments already using capable models for attacks, fraud, and surveillance.
- 1AI will intensify a cyber arms race as attackers and defenders deploy increasingly capable systems.
- 2Sardine uses proprietary consortium data and machine-speed agents to reduce human fraud review dramatically.
- 3Autonomous security tools lower the cost of expert work while raising new questions about jobs and workplace privacy.
Don't miss
Kevin Mandia says complex web-application assessments have fallen from several days of human work to roughly ten minutes and a few dollars.
The brief
Kevin Mandia and Supes Ranjan join the discussion as the episode shifts AI safety away from science fiction and toward criminals, companies, and nation-states using models today.
Mandia argues that organizations should prepare for an AI cyber arms race regardless of whether frontier labs slow development, because attackers and defenders will both adopt capable tools.
Supes Ranjan describes Sardine’s machine-speed fraud investigations, powered by proprietary consortium data, while the panel weighs open models against expensive closed systems.
The sharpest cost shock comes in cybersecurity: complex web-application assessments that once took days can now take roughly ten minutes and a few dollars.
AI may augment workers by lifting beginners toward intermediate skill, but it could also reduce junior security roles and make recorded workplaces far less private.