LMDeploy is a toolkit for compressing, deploying, and serving large language models. Versions prior to 0.12.3 have a Server-Side Request Forgery (SSRF) vulnerability in LMDeploy's vision-language module. The load_image() function in lmdeploy/vl/utils.py fetches arbitrary URLs without validating internal/private IP addresses, allowing attackers to access cloud metadata services, internal networks, and sensitive resources. Version 0.12.3 patches the issue.
{
"osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/33xxx/CVE-2026-33626.json",
"cna_assigner": "GitHub_M",
"cwe_ids": [
"CWE-918"
]
}{
"source": [
"CPE_RANGE",
"REFERENCES"
],
"cpe": "cpe:2.3:a:internlm:lmdeploy:*:*:*:*:*:*:*:*",
"extracted_events": [
{
"introduced": "0"
},
{
"fixed": "0.12.3"
}
]
}