Everyday AI Podcast – An AI and ChatGPT Podcast Ep 718: Agent Risk, Security, and AI Sprawl in 2026: Why AI That Acts Changes Everything (Start Here Series Vol 9)
74 snips
Feb 20, 2026 A deep dive into how AI that actually acts creates new security and sprawl problems. They trace the evolution from chatbots to proactive agents and map three risk surfaces: inputs, tools, and actions. Shadow AI and dark agent proliferation get flagged as a hidden storm. Practical responses and lab tradeoffs, plus supply-chain and browser agent dangers, are highlighted.
AI Snips
Chapters
Transcript
Episode notes
Host Runs Multiple Agents Locally
- Jordan Wilson describes running multiple agents (CodeX, cloud code, anti-gravity) on his personal computer.
- He inspects results afterward but admits he doesn't always know what agents do between runs.
Three Shades Of Dark AI
- Dark AI emerges in three forms: shadow AI, agent sprawl, and dark agent sprawl with unknown agents.
- Jordan Wilson warns dark agent sprawl can include malicious agent malware or invisible unapproved agents in companies.
The Perfect Storm: Reasoning, Use, Memory
- Three forces created the agentic perfect storm: reasoning, computer use, and long-context memory.
- Jordan Wilson links model reasoning and UI/computer control improvements to rapid, hard-to-see agent growth.