@oreillymedia: It might sound like a naive security check, but bad actors regularly hide malicious instructions inside hidden Unicode characters in AI agent files. Maxim Salnikov explains how APM scans for these characters automatically, and why enterprise teams should pair that scan with policy governance to keep agent customizations safe at scale.

O’Reilly

Open In TikTok:

Region: US

Tuesday 20 October 2026 17:04:01 GMT