IBM Cloud Pak for Business Automation CVE-2026-12751

Exploitation requires a low-privileged authenticated account on the IBM Cloud Pak for Business Automation web UI (CVSS PR:L) and a reachable page or field that reflects or stores attacker-supplied HTML without script-tag or attribute neutralization (CWE-80). A victim must then view the injected content (UI:R) in their browser, and because the vector carries S:C, the injected script executes in the security context of the hosting site's origin rather than a separate origin. What limits exploitation: the attacker cannot trigger execution anonymously - credentials are needed to inject - and nothing executes until a second user loads the poisoned page, so successful attacks are usually targeted at specific users or roles rather than mass-triggered. No CISA KEV listing, EPSS score, or public proof-of-concept was included in the feed, so confirmed in-the-wild exploitation cannot be asserted, and the exact non-default configuration or endpoint that is vulnerable is not described in the input and must be taken from the IBM advisory.

Multiple signals point to a moderate, real-but-bounded risk. The CVSS vector shows a network attack vector and low attack complexity, which raises the ceiling, but PR:L and UI:R pull the score down to 5.4 - the attacker must be an authenticated low-privileged user and a victim must view the crafted content, so this is not a drive-by unauthenticated compromise. S:C and the C:L/I:L impacts are consistent with classic script injection in the site's origin, which is meaningful for session and credential theft but does not imply server-side code execution or data destruction (A:N). No EPSS score was supplied, no CISA KEV listing is present, and no public proof-of-concept is indicated, so there is no evidence of active or widespread exploitation; the SSVC-style reading would be 'PoC/exploitation status unknown, not automatable (requires authenticated delivery plus victim interaction), technical impact partial.' Key data gaps: the CPE is wildcarded so the affected version range is unconfirmed, and the advisory reference points to an IBM Support page whose fix version is not enumerated in the input - verify affected releases directly against that page before triaging exposure.

An attacker with valid low-privileged credentials on a Cloud Pak for Business Automation console submits crafted HTML, for example through a stored field, comment, or reflected request parameter, that the application renders without encoding. When a higher-privileged user or administrator later opens the affected page, the markup executes in the browser under the application's origin, letting the attacker read session tokens, act on the victim's behalf, or exfiltrate data visible to that user. Exploitation requires both the authenticated delivery step and the victim's page view, so it is typically mounted in phishing or insider-style scenarios rather than as an internet-wide automated scan, and no public exploit code was identified in the supplied data.