
AI Security Industry Overview

Why securing AI is harder than anyone expected and the coming security crisis | Sander Schulhoff
Topic 8 of 14
Published Dec 21, 2025 · Clip 19:45–23:56 (4:11)
Sander differentiates frontier labs' research from B2B AI security vendors offering monitoring, compliance, red teaming, and guardrails. He explains automated red teaming uses LLMs to generate malicious prompts for testing. Guardrails classify inputs/outputs, deployed before/after the main model to flag harms like hate speech or CBRNE info.