Why won’t AI agents just follow the rules?

Episode 50 of Security Intelligence discusses why AI agents deviate from rules, drawing on the HuggingFace hack and a Dark Reading op-ed to explore AI ethics. Dustin “EvilMog” Heywood, Seth Glasgow, and Nick Bradley join host Matt Kosinski to examine balancing probabilistic AI utility with deterministic controls. The podcast also covers the OWASP Top 10 for agentic skills, the impact of AI on bug bounty programs, and introduces ThreatXtension for detecting malicious browser extensions.