Maximo AI Service CVE-2026-90912

No special configuration conditions are stated: per CVSS AV:N/AC:L/PR:N/UI:N, a remote unauthenticated attacker can reach the affected processing path in IBM Maximo AI Service SaaS versions 9.1.0 and 9.2.0 and supply input that triggers uncontrolled recursion. The primary limiting factors are (1) the product is delivered as SaaS, so the exposed instance and its patching are controlled by IBM, narrowing the realistic attacker set to those who can reach a given tenant's endpoint; (2) only the 9.1.0 and 9.2.0 service levels are identified as affected; and (3) impact is limited to availability - no data disclosure or integrity loss results from a successful trigger.

The input signals are moderately consistent but incomplete. The CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H, base 7.5) describes a remotely reachable, low-complexity, unauthenticated availability-only impact, which matches the CWE-674 (uncontrolled recursion) root cause and the stated denial-of-service outcome - these signals do not conflict. However, several decision-driving signals are absent: no EPSS probability was supplied, there is no CISA KEV listing, and no public exploit code is referenced, so the only exploitation evidence is the theoretical reachability implied by the CVSS vector. The impact is confined to availability (C:N/I:N) with no scope change (S:U), so this is not a data-breach or code-execution issue - it is a crash/hang/resource-exhaustion condition. Because the affected product is delivered as SaaS (CPE cpe:2.3:a:ibm:maximo_ai_service), the attack surface and patching are largely controlled by IBM, which meaningfully reduces real-world exposure for customers compared with on-prem software carrying the same score. Net assessment: this is a legitimate but mid-tier priority - a high-CVSS availability flaw with a vendor patch already available and no evidence of active or public exploitation. EPSS is not provided, so no exploitation-probability comparison can be made; treat the 7.5 as an upper bound on urgency rather than confirmation of imminent threat.

A remote, unauthenticated attacker sends a specially crafted request or payload to the Maximo AI Service endpoint whose structure drives the vulnerable parser or processing routine into unbounded recursion. The resulting stack exhaustion or runaway resource consumption crashes or hangs the service, denying availability to legitimate users. No public exploit code is referenced, so exploitation would require an attacker to independently craft the recursive input; given AC:L the trigger is likely straightforward once the vulnerable input path is known.