An AI coding agent deleted PocketOS's production database and all backups in 9 seconds by issuing a single unconfirmed API command—but the real failure was infrastructure design: the backups were stored in the same volume as production data, so they got wiped too. This post cuts through the "rogue AI" narrative to point out that news coverage is blaming the AI agent while overlooking the systemic problem—Railway's backup architecture should never have allowed this, and the company should have had geographically separated or immutable backups. The post doesn't blame the founder (who made a genuine error in tooling choice) but rather the framing that treats AI as the villain when infrastructure negligence was the actual root cause.