AI-Built Apps Are Leaking Private Company Data

Researchers just found thousands of AI-built apps leaking medical records, financial data, and customer PII straight to the open internet. The scary part isn't that AI writes code — it's that it writes code just well enough that nobody asks questions. Frank Downs and Dustin Brewer break down the hidden cost of vibe coding: insecure-by-default software shipped to production, AI tools replacing the junior developers who'd grow into the people who fix it, and AI quietly wired into services you never consented to — including a dentist's chair that records every cleaning and sends it to an insurance-linked system. AI learned security from us. And we were never good at it. 🎙️ Listen: https://legitimatecybersecurity.podbean.com/ 📩 Media/interview: [email protected] Hosted by Frank Downs and Dustin Brewer. Chapters: 00:00 The code works — that's the problem 01:24 "Do you consider yourself a coder?" 03:15 What AI actually learned to copy (us) 04:58 Vibe-coded tools running in production 05:19 3,380 exposed apps, 5,000 data leaks 07:56 Who fixes it when the cyber team finds holes? 08:26 The $1.5M QA cut that cost $6M 09:35 AI talking to AI: nobody reads the code 15:21 "Your password is God" — security never changed 16:27 Should AI touch the live service? 17:48 The dentist chair that records everything 21:00 Where the line actually is (help desk vs. prod) 24:20 AI monitoring employees & the gold-standard trap 28:23 Always-on "streaming AI" is 5 years out 29:25 The coming AI caste system 30:34 Adversaries already use it (the Lego propaganda) 33:14 We're about to lose every junior analyst 40:15 The Twitter "efficiency" parallel 41:35 Keep on cybering #vibecoding #cybersecurity #aisecurity #dataprivacy #shadowit #infosec #aitools #privacy #devsecops #surveillance