A security flaw in the Home Assistant Green smart-home hub can let an attacker who has already gained a foothold inside a container read user account tokens stored in plaintext and escalate to full control of the device, researchers have disclosed.

The flaw, tracked as ZDI-26-896 and published by the Zero Day Initiative on 6 November, carries a CVSS severity score of 8.8 out of 10. It was found by Team ANHTUD during the Pwn2Own Ireland hacking contest and is listed under CVE-2026-91131.

According to the advisory, the hub stores credentials in cleartext. Low-privileged code running inside a container can read them, escape the container and execute commands as root on the host system.

The attack needs an initial foothold: the attacker must first be able to run low-privileged code inside a container on the device. It is not remotely exploitable on its own, but it turns a limited compromise into a full takeover of a device that controls lights, locks, heating and cameras.

Home Assistant has shipped an update addressing the issue, and owners should install the latest release. The disclosure came alongside a companion advisory, ZDI-26-897, describing a server-side request forgery flaw in the same device. Users should also review which add-ons and integrations can run code on their hub.