watsonx.data Intelligence CVE-2025-36333

Exploitation requires a valid authenticated session with at minimum low-privilege user credentials on IBM watsonx.data Intelligence (confirmed by CVSS PR:L). The vulnerability is triggered through the platform's behavioral workflow engine - no special non-default configuration or specific feature flag is documented as required beyond having an active authenticated session. Limiting factors include the mandatory authentication barrier (unauthenticated remote exploitation is not possible), the confined integrity-only impact (C:N, A:N limits data exfiltration and denial-of-service scenarios), and the product's enterprise-only deployment context. The vulnerability does not affect watsonx.data versions outside 5.2.0-5.3.0.

The CVSS 3.1 base score of 4.3 (Medium) reflects limited real-world impact: attack vector is network (AV:N), complexity is low (AC:L), low privileges are required (PR:L), no user interaction is needed (UI:N), scope is unchanged (S:U), and impact is confined to integrity only (I:L, C:N, A:N). This profile describes a meaningful but bounded risk - any authenticated user could exploit it without needing admin rights or special configuration, lowering the practical bar. However, the absence of confidentiality and availability impact significantly constrains the blast radius. No EPSS data was provided in the input. The vulnerability is not in the CISA KEV catalog and no public proof-of-concept exploit has been identified, both of which reduce the urgency of emergency patching. The primary real-world risk is insider threat or compromised low-privilege credentials enabling privilege-of-action escalation within the platform workflow. This is a genuine medium-priority issue for organizations running watsonx.data Intelligence, not a critical emergency.

An authenticated low-privilege user with a valid session to the IBM watsonx.data Intelligence web interface crafts or replays API requests that invoke workflow actions out of their intended sequence, bypassing server-side state validation and executing operations they are not authorized to perform. No public proof-of-concept exploit has been identified, but the low complexity and network accessibility mean any credentialed internal user or attacker with stolen credentials could attempt this without specialized tooling.