{"data":{"id":"402bf3d7-d8a5-410b-b4e7-d9b4f780eea2","title":"CVE-2026-55646: vLLM is an inference and serving engine for large language models. From 0.22.0 to 0.23.0, the /v1/audio/transcriptions…","summary":"From vLLM 0.22.0 to 0.23.0, the /v1/audio/transcriptions and /v1/audio/translations routes call request.file.read() and fully load an uploaded audio file into memory before the VLLM_MAX_AUDIO_CLIP_FILESIZE_MB size limit (default 25 MB) is checked during speech-to-text preprocessing. An API caller who can reach these routes can submit an oversized multipart upload, causing memory allocation proportional to the file size before rejection, which can create memory pressure or terminate the process depending on deployment resource limits.","solution":"This issue is fixed in version 0.24.0.","labels":["security"],"sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-55646","publishedAt":"2026-07-06T20:16:37.663Z","cveId":"CVE-2026-55646","cweIds":["CWE-400","CWE-770"],"cvssScore":"6.5","cvssSeverity":"medium","severity":"medium","attackType":["denial_of_service"],"issueType":"vulnerability","affectedPackages":["vllm@>= 0.22.0, < 0.24.0 (fixed: 0.24.0)"],"affectedPackageNames":["vllm"],"affectedPackageRefs":["pypi:vllm"],"affectedVendors":[],"affectedVendorsRaw":["vLLM"],"classifierModel":"claude-haiku-5-5","classifierPromptVersion":"v4","summaryPromptVersion":"v2","headline":"vLLM audio transcription routes allow memory exhaustion via oversized uploads","headlinePromptVersion":"h1","cvssVector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H","attackVector":"network","attackComplexity":"low","privilegesRequired":"low","userInteraction":"none","exploitMaturity":"unknown","epssScore":0.00519,"epssCheckedAt":"2026-10-14T06:09:06.335Z","kevDateAdded":null,"advisoryAliases":["GHSA-v82g-2437-67m2"],"affectedPackagesSource":"ghsa","affectedPackagesCheckedAt":"2026-10-10T03:47:01.551Z","patchAvailable":true,"disclosureDate":"2026-07-06T20:16:37.663Z","capecIds":["CAPEC-125","CAPEC-130"],"crossRefCount":0,"attackSophistication":"trivial","impactType":["availability"],"aiComponentTargeted":"api","llmSpecific":true,"classifierConfidence":0.93,"researchCategory":null,"atlasIds":null}}