vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion. The AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions in multimodal/inputs.py fetch user-supplied media URLs using aiohttp and call r.read() without enforcing a maximum response size, allowing an attacker to exhaust server memory by providing a URL to an arbitrarily large file.

CVE-2026-37237 affects vLLM versions up to and including 0.17.0, allowing remote attackers to cause a Denial of Service through memory exhaustion. The vulnerability resides in the AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions within multimodal/inputs.py. These functions fetch user-supplied media URLs using aiohttp and invoke r.read() without enforcing any maximum response size limit. An attacker can exploit this by supplying a URL pointing to an arbitrarily large file, causing the server to read and buffer the entire response into memory. This can result in complete memory exhaustion on the affected server, leading to service disruption. A fix has been proposed via GitHub pull request #36506 against the vllm-project repository.