CVE-2026-37237: vLLM Unbounded Media Fetching Memory Exhaustion
vLLM
vLLM multimodal input fetching functions fail to enforce maximum response sizes when reading user-supplied media URLs. Remote attackers can exploit this lack of limits to exhaust server memory and cause denial of service.
disclosed CVENo confirmed exploitation, so this record is tracked as a disclosed vulnerability and not counted as an incident. It moves into the counts if CISA or the vendor confirms exploitation. Why
- CVSS
- 7.5
- EPSS
- 0.01 · top 46% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 28 Aug 2026
- last update
- 28 Aug 2026
- CVE ids
Timeline · 1
Sources · 1
Advisories and databases
- 2026-08-28CVE-2026-37237: vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion.NVD · nvd.nist.gov/vuln/detail/CVE-2026-37237
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerability
- severity
- 2 · Disclosed vulnerability
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- no technique for these categories
- dates
- happened 28 Aug 2026 · first reported 28 Aug 2026
- origin
- NVDVulnerability record from NVD / the GitHub Advisory Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- record
- confidence 0.95 · created 2026-10-18 06:33Z · updated 2026-10-18 06:33Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# CVE-2026-37237: vLLM Unbounded Media Fetching Memory Exhaustion https://gcdradar.vercel.app/incidents/cve-2026-37237-vllm-unbounded-media-fetching-memory-exhausti · AIR-2026-0096 · closed 2026-08-28 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2026-08-28 - AI role: Attack on AI - Products: vLLM - CVE ids: CVE-2026-37237 - CVSS: 7.5 ## Summary vLLM multimodal input fetching functions fail to enforce maximum response sizes when reading user-supplied media URLs. Remote attackers can exploit this lack of limits to exhaust server memory and cause denial of service. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2026-08-28: CVE-2026-37237: vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion. (https://nvd.nist.gov/vuln/detail/CVE-2026-37237) ## Sources - 2026-08-28 NVD: CVE-2026-37237: vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion. https://nvd.nist.gov/vuln/detail/CVE-2026-37237 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closed6.9vLLM vLLM Audio Size Enforcement Bypass Permits Resource Exhaustion
Patch now · Disclosed · since 26 Sep 2026 · CVE-2026-100648
AI - closed6.3vLLM vLLM Decoder Allocation Resource Limit Bypass Causes GPU Memory Exhaustion
Patch now · Disclosed · since 26 Sep 2026 · CVE-2026-100649
AI - closed8.2vLLM vLLM Stop Token ID Boundary Validation Flaw Causes Service Crash
Patch now · Disclosed · since 26 Sep 2026 · CVE-2026-100652
AI - closed6.9vLLM CVE-2026-100647: vLLM denial-of-service vulnerability in cache_salt parameter
Patch now · Disclosed · since 26 Sep 2026 · CVE-2026-100647
AI - closed8.3vLLM CVE-2026-100653: vLLM revision pin failure during Hugging Face artifact loads
Patch now · Disclosed · since 26 Sep 2026 · CVE-2026-100653
AI - closed7.1vLLM CVE-2026-100651: vLLM omits prompt-length validation on disaggregated endpoint
Patch now · Disclosed · since 26 Sep 2026 · CVE-2026-100651
AI