AI Didn’t Go Rogue—Your Permissions Did

Send us Fan Mail In April 2026, an AI coding agent deleted a company’s entire production database—in just nine seconds. The backups disappeared too. It sounds like a story about AI going rogue. But the real problem wasn’t the AI itself. A forgotten access token had far more permissions than it needed, destructive actions didn’t require human approval, and backups weren’t sufficiently isolated from the production environment. In this episode, we unpack what happened at PocketOS and what the incident reveals about a much larger challenge facing businesses adopting AI: **AI...**