Vulnerability  ·  2026-09-08

JeecgBoot AIRAG LLM-model export access-control flaw allows credential-argument manipulation

VulnerabilityMedium impactGlobalCVE-2026-86228
A CVE record published 2026-09-06 documents improper access control in JeecgBoot's AiragModelController.java exportXls function within the airag/llm module (versions up to 3.9.3). An attacker can manipulate the credential argument to access information through the AI-model export route; the record notes public exploit disclosure. CVSS v3.1 4.3 / CVSS v4.0 5.3.
JeecgBoot is a widely-used low-code development platform bundling an AIRAG (AI + RAG) module; exposure of credentials or model-export data through a broken authorization check on an LLM-integrated export function could leak API keys or internal model configuration data tied to the RAG pipeline, though the narrow scope (single export function, medium severity, single vendor) limits broader blast radius.
A remote attacker manipulates the 'credential' argument passed to the exportXls function in the AI/RAG model-export controller, bypassing intended access control to gain unauthorized access to information exposed via the LLM-model export path.
JeecgBoot ≤ 3.9.3 (jeecg-boot-module-airag component, AiragModelController.java exportXls function)
Upgrade to JeecgBoot 3.9.5 (patched); review AI-model credentials and exports; enforce role-based authorization checks on all AIRAG/LLM endpoints.
VulDB CVE-2026-86228CVE.org Record
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →