CVE-2026-41713: Prompt Injection via Memory Poisoning in PromptChatMemoryAdvisor
A malicious user could craft input that is stored in conversation memory and later interpreted by the model in an unintended way. Applications using the affected advisor with user-controlled input may be susceptible to manipulation of model behavior across conversation turns.
Affected Software
Event History
Frequently Asked Questions
What is the severity of CVE-2026-41713?
CVE-2026-41713 has been assigned a severity rating that indicates it poses a risk of prompt injection via memory poisoning.
How do I fix CVE-2026-41713?
To mitigate CVE-2026-41713, update to a version of VMware Spring Ai that is not affected, specifically versions 1.0.8 or later for 1.0.x and 1.1.7 or later for 1.1.x.
Which versions of VMware Spring Ai are impacted by CVE-2026-41713?
CVE-2026-41713 affects VMware Spring Ai versions from 1.0.0 to 1.0.7 and versions from 1.1.0 to 1.1.6.
What kind of exploitation is possible with CVE-2026-41713?
CVE-2026-41713 allows a malicious user to perform prompt injection through memory manipulation, which could lead to unintended model behavior.
Is user-controlled input a factor in CVE-2026-41713?
Yes, user-controlled input is a critical factor in CVE-2026-41713 as it can be exploited to store and manipulate data within conversation memory.