ai-fusion-video EUVD-2026-116897

Requires a valid authenticated account on the target ai-fusion-video instance (PR:L), but any ordinary registered low-privileged user suffices - no admin role, no special configuration, and no user interaction are needed. The concrete trigger is supplying another user's unscoped, sequential auto-increment object id to the /api/project, /api/asset or /api/script REST handlers, which fail to check object ownership. All default deployments of versions before 1.3.0 are affected. The only meaningful limiting factor is the need for authentication, which keeps this out of reach of fully anonymous attackers but not of any legitimate or self-registered user.

Signals are largely consistent and point to a genuine but not emergency-grade priority. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) describes a network-reachable, low-complexity flaw needing no user interaction, and the PR:L confirms a low-privileged authenticated account is required - this is broken object-level authorization (CWE-639), not an unauthenticated authentication bypass despite the input tag. Impact is high across confidentiality, integrity and availability (VC:H/VI:H/VA:H) because the description shows attackers can read, overwrite, and cascade-delete other tenants' data and add themselves as project members, which the SC:N/SI:N/SA:N values correctly keep scoped to the application itself. No EPSS probability and no CISA KEV entry were provided in the input, so there is no evidence of active or widespread exploitation; the vuln.today Decision label is 'this_week' rather than act_now/emergency, consistent with a serious issue that should be remediated promptly but is not under confirmed attack. A public PoC reference exists (a HackMD write-up and the GitHub issue) but it is not verified as working exploit code, so POC availability must not be conflated with active exploitation. The main factor tempering raw severity is the authentication requirement (any registered user, but not fully anonymous); the main factor elevating it is that the trigger - enumerating sequential auto-increment IDs - is trivial for any logged-in user in a multi-tenant deployment.

A low-privileged user registers or already holds an account on an ai-fusion-video instance, then iterates sequential auto-increment IDs against the /api/project, /api/asset and /api/script handlers to reach other users' objects. Because the handlers do not verify object ownership, the attacker reads confidential project data, overwrites or cascade-deletes victim records, and silently adds their own account as a member of victim projects. A public PoC reference (HackMD write-up) describes the technique, though it is not verified as working exploit code; the network vector and low attack complexity make the attack straightforward once authenticated.