This vulnerability (CVE-2026-41713) affects VMware Spring AI versions 1.0.0 and 1.1.0 and is classified under CWE-1336, which involves improper neutralization of special elements in a template engine. An attacker can craft input that is stored in the conversation memory and later interpreted by the AI model in an unintended way, potentially manipulating the model's behavior across multiple conversation turns. The CVSS 3.1 vector indicates the attack can be performed remotely without privileges or user interaction, with low attack complexity. The impact includes low confidentiality loss but high integrity impact, with no effect on availability. No official remediation or patch is currently documented by VMware, and this is not a cloud service vulnerability.

The vulnerability allows an attacker to manipulate the behavior of the AI model by injecting crafted input that is stored and later interpreted incorrectly. This can lead to unauthorized modification of model behavior, impacting the integrity of the system's responses. Confidentiality impact is low, and availability is not affected. There are no known exploits in the wild, so active exploitation has not been observed.

Patch status is not yet confirmed — check the VMware advisory for current remediation guidance. Since no official patch or temporary fix is documented, users should monitor VMware communications for updates. Until a fix is available, restricting or sanitizing user-controlled input before it is stored in conversation memory may reduce risk, but this is not a vendor-provided mitigation.