AI Engine EUVD-2026-103398

Requires an authenticated WordPress account with the Editor role or higher on a site running AI Engine before 3.7.9; the attacker then calls one of the plugin's media actions with the ID of an object they do not own. No special plugin configuration beyond having the vulnerable version active is described - the missing ownership/authorization check is present by default. Exploitation is strictly limited by the trusted-role requirement (Editor+), which prevents anonymous or low-privilege (Subscriber/Contributor) abuse and confines realistic risk to insider or compromised-account scenarios; no user interaction by a victim is needed.

This is a genuine but moderate-priority issue, not a critical internet-facing emergency. The CVSS 3.1 vector (CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:N/I:H/A:N, base 4.9) correctly captures the dominant limiting factor: exploitation requires an authenticated account with Editor role or above (PR:H/PR:L), so there is no path for an anonymous remote attacker. The CWE-862 (Missing Authorization) classification and description are consistent - this is a broken-access-control / insecure-direct-object-reference flaw, NOT an authentication bypass despite the misleading input tag. POC/exploit code is publicly available via WPScan, so weaponization is trivial once a sufficiently privileged account exists, but there is no CISA KEV listing and no EPSS score was supplied, so there is no evidence of active or widespread exploitation. Signals are internally consistent except for privilege level: the input vector uses PR:H, yet the description grants the capability to the Editor role, which is typically scored PR:L in WordPress - this discrepancy understates reach slightly (verify role mapping). Net: real-world risk is dominated by insider or compromised-privileged-account scenarios rather than mass exploitation, making this a timely-patch item rather than a drop-everything emergency.

A disgruntled or compromised user holding an Editor account on a WordPress site running AI Engine < 3.7.9 invokes one of the plugin's media actions, supplying the ID of a post object they do not own. Because the plugin never verifies ownership before converting the item to an attachment and permanently deleting it, the attacker destroys protected content - navigation menus, block templates, or global styles - that their role should not be able to touch. Publicly available exploit code from WPScan lowers the effort required to carry this out.