IBM AIX CVE-2026-18835

Exploitation requires the attacker to already hold valid low-privilege credentials and network reachability to the affected AIX or VIOS management/command interface (CVSS PR:L, AV:N, AC:L, UI:N). The concrete prerequisite is authenticated access to the component that improperly neutralizes OS-command special elements on AIX 7.2/7.3 or VIOS 4.1; the specific feature, daemon, or CLI is not named in the advisory data provided. Limiting factors: the authentication requirement blocks fully unauthenticated attackers, and exploitation is confined to the affected AIX/VIOS versions on IBM POWER hardware, so environments that isolate management interfaces or restrict account provisioning materially reduce exposure.

The signals are largely aligned toward high priority: the CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:C, all impacts High) yields a 9.9 base score driven by network reach, low attack complexity, only low privileges required, no user interaction, and a scope change into the full OS. That said, PR:L means an attacker must first hold a valid (even unprivileged) account, which is a meaningful gate against fully anonymous internet-wide exploitation and argues for targeted rather than mass exploitation. Counterbalancing the high CVSS: there is no EPSS score, no POC, and no KEV listing in the provided data, so real-world exploitation probability is unquantified and currently unproven. This is a genuine high-severity issue warranting prompt patching on management/administrative planes, but the near-term likelihood of opportunistic mass exploitation is unconfirmed given the authentication requirement and absence of public exploit code.

An attacker who has obtained a low-privilege authenticated session on an exposed AIX 7.2/7.3 host or a VIOS 4.1 management interface submits input containing shell metacharacters to the vulnerable command-processing function. Because the input is not neutralized, the injected commands execute on the underlying OS, and the CVSS scope change indicates the attacker can move beyond the original component to compromise the full host. No public POC is currently identified, so exploitation would require the attacker to independently identify the vulnerable sink.